[rust] Support building and running the Rust protobuf upb kernel on wasm32-unknown-emscripten

### Motivation

While Google3 defaults `--//third_party/protobuf/rust:rust_proto_library_kernel`
to `cpp`, the open-source `protobuf` crate ships only the `upb` kernel
(`third_party/protobuf/rust/release_crates/google_protobuf/build.rs` compiles
`libupb`), and Wasm (`wasm32-unknown-emscripten`) targets that want to evaluate
or build Rust protobufs with the `upb` kernel currently fail at three build
layers plus link time.

### The 4 Blockers on `wasm32-unknown-emscripten` and Their Fixes

1. **Blaze `select()` conflict between `:use_upb_kernel` and `force_lite_runtime`**
   - **Exact error (`wasm_upb_try1.log`)**:
     ```
     ERROR: third_party/protobuf/rust/BUILD:80:13: Illegal ambiguous match on
     configurable attribute "deps" in //third_party/protobuf/rust:protobuf:
     //third_party/protobuf/rust:use_upb_kernel
     //net/proto2/public:force_lite_runtime
     Multiple matches are not allowed unless one is unambiguously more specialized.
     ```
   - **Why `--features=-proto_force_lite_runtime` cannot work around it (`wasm_upb_try2.log`)**:
     ```
     ERROR: The C++ toolchain '//third_party/cpptoolchains/portable_llvm/stable:cc_toolchain_wasm'
     unconditionally implies feature 'proto_force_lite_runtime', which is disabled by command line
     ```
   - **Fix (`third_party/protobuf/rust/BUILD`)**:
     - `deps`: add a `:use_cpp_kernel` `config_setting`
       (`rust_proto_library_kernel = "cpp"`) and a
       `selects.config_setting_group(name = "force_lite_runtime_cpp_kernel",
       match_all = [":use_cpp_kernel", "//net/proto2/public:force_lite_runtime"])`,
       replacing `//net/proto2/public:force_lite_runtime` in the `deps` `select()`.
     - `rustc_flags` of `:protobuf` and `:protobuf_lite`: split into two
       concatenated, orthogonal `select()`s (kernel: `--cfg=upb_kernel` /
       `--cfg=cpp_kernel`; `//net/proto2/public:force_lite_runtime`:
       `--cfg=lite_runtime`). cpp and cpp + force-lite get exactly the same flags
       as before; upb + force-lite now also gets `--cfg=lite_runtime`, gating off the
       `not(lite_runtime)` items in `protobuf.rs` (so it reduces to
       `pub use protobuf_lite::*`) to match the lite gencode that `aspects.bzl`
       emits under `proto_force_lite_runtime` for both kernels.

2. **`linkme::distributed_slice` unsupported on `target_os = "emscripten"`**
   - **Exact error (`wasm_upb_try3.log`)**:
     ```
     error: distributed_slice is not implemented for this platform
       --> third_party/protobuf/rust/upb_kernel/extension.rs:40:1
        |
     40 | #[linkme::distributed_slice]
        | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
     error[E0425]: cannot find value `LINKME_START_EXTENSIONS` in this scope
     error[E0425]: cannot find value `LINKME_STOP_EXTENSIONS` in this scope
     error[E0425]: cannot find value `LINKME_DUPCHECK_START_EXTENSIONS` in this scope
     error[E0425]: cannot find value `LINKME_DUPCHECK_STOP_EXTENSIONS` in this scope
     ```
   - **Root cause**: `linkme` 0.3.35 has no `target_os = "emscripten"` backend
     (the per-OS `link_section` `cfg_attr` list in `third_party/rust/linkme_impl/v0_3/src/declaration.rs`), and
     cannot be extended directly: on Wasm targets (`is_like_wasm`), `rustc` emits
     `#[link_section]` statics as non-addressable Wasm custom sections, with a
     hardcoded exception only for `.init_array*` sections
     (`rustc_codegen_llvm/src/consts.rs:545-560`).
   - **Fix (`third_party/protobuf/rust/upb_kernel/extension.rs`)**: Gate
     `#[linkme::distributed_slice]` on `#[cfg(not(target_os = "emscripten"))]`. On
     `#[cfg(target_os = "emscripten")]`:
     - Provide an empty `pub static EXTENSIONS: linkme::DistributedSlice<[LazyLock<MiniTableExtensionInitPtr>]>`
       in the value namespace so generated `#[linkme::distributed_slice($pbr$::EXTENSIONS)]`
       attributes continue to type-check without modifying `protoc`'s Rust code
       generator (`compiler/rust/extension.cc`).
     - Export `#[macro_export] macro_rules! __protobuf_upb_emscripten_extensions`
       re-exported as `pub use __protobuf_upb_emscripten_extensions as EXTENSIONS;`
       in the macro namespace. `#[linkme::distributed_slice($pbr$::EXTENSIONS)]`
       expands to `$pbr$::EXTENSIONS! { #![linkme_macro = ...] pub static ... }`
       (`third_party/rust/linkme_impl/v0_3/src/element.rs`, `expand`), which
       our macro intercepts to emit the extension's `LazyLock<MiniTableExtensionInitPtr>`
       static alongside a `#[used] #[unsafe(link_section = ".init_array.101")]`
       constructor that pushes `&'static LazyLock<MiniTableExtensionInitPtr>` into
       `EMSCRIPTEN_EXTENSIONS` during `__wasm_call_ctors` (`_initialize`).
     - In `generated_extension_registry()`, iterate `EMSCRIPTEN_EXTENSIONS` and
       register each extension in the `upb_ExtensionRegistry`.

3. **Unstable `core::arch::global_asm!` on `wasm32`**
   - **Exact error (`wasm_upb_try3.log`)**:
     ```
     error[E0658]: inline assembly is not stable yet on this architecture
       --> third_party/protobuf/rust/upb_kernel/extension.rs:44:1
        |
     44 | / core::arch::global_asm!(
     ...  |
     55 | | );
        | |_^
     ```
   - **Root cause**: `wasm32-unknown-emscripten` sets `target_family = "unix"`,
     so it matches the ELF sentinel `global_asm!` gate.
   - **Fix (`third_party/protobuf/rust/upb_kernel/extension.rs`)**: Add
     `not(target_os = "emscripten")` to both `global_asm!` gates (head already
     excludes `target_vendor = "apple"`).

4. **Emscripten link failure when `longjmp = "off"`**
   - **Exact error (`wasm_upb_try4.log`)**:
     ```
     emcc: error: '.../wasm-ld ...':
     wasm-ld: error: ...: undefined symbol: __wasm_longjmp
     error: longjmp support was disabled (SUPPORT_LONGJMP=0), but it is required by the code (either set SUPPORT_LONGJMP=1, or remove uses of it in the project)
     ```
   - **Root cause**: `upb`'s wire decoder (`upb_Decode`), wire encoder
     (`upb_Encode`), and MiniTable builder (`upb_MiniTable_Build`) use
     `UPB_SETJMP` / `UPB_LONGJMP` (`setjmp` / `longjmp` in
     `third_party/upb/upb/port/def.inc`) for error unwinding.
   - **Requirement (documented in `third_party/protobuf/rust/test/upb/BUILD` and
     `wasm_extension_test.rs`)**: `wasm_cc_binary` targets linking `upb` must set
     `longjmp = "wasm"` (Wasm Exception Handling instructions, required when
     `standalone = True`), use the default `longjmp = "emscripten"`, or compile
     `upb` with `--copt=-DWASM_WAMR=1` (which replaces `UPB_SETJMP`/`UPB_LONGJMP`
     with `0`/`abort()`).

### Caveats

- **Unreferenced extension-only crates on `wasm32`**: On `wasm32`, `rustc`
  sets `has_symbols_obj = false` (`Target::object_architecture` returns `None`,
  so `add_linked_symbol_object` in `rustc_codegen_ssa/src/back/link.rs` emits no
  `symbols.o` table of `#[used]` statics into `.rlib` archives). Consequently,
  when `wasm-ld` links an `.rlib` archive without `--whole-archive`, an object
  file containing an `.init_array` extension constructor is extracted only if at
  least one symbol from that codegen unit is referenced by the binary. Extensions
  in crates that are never referenced by any code path will not have their
  `.init_array` constructors linked on Wasm (whereas native `linkme` retains them
  via `symbols.o`).
- **Constructor priority `.init_array.101` + late-registration check**:
  Bare `.init_array` entries run at default priority `65535`. If another static
  constructor at default priority parses a message before all extension
  constructors have run, `generated_extension_registry()` would freeze the
  `LazyLock` registry early. The macro therefore places constructors in
  `.init_array.101`, and `__emscripten_register_extension` asserts via
  `debug_assert!` that `EMSCRIPTEN_REGISTRY_BUILT` is still `false`.
- **Copybara compatibility**: `load("//third_party/bazel_skylib/lib:selects.bzl", "selects")`
  in `third_party/protobuf/rust/BUILD` and the Google3-only Wasm targets/loads in
  `third_party/protobuf/rust/test/upb/BUILD` (`wasm_cc_binary`, `build_test`,
  `rust_binary`) are annotated with `# copybara:strip` /
  `# copybara:strip_begin` / `# copybara:strip_end_and_replace_begin` so OSS
  Bazel exports remain clean.
- **Second linkme slice (`EXTENSION_MINI_TABLE_TO_DEF_INITS`, reflection)**: both
  linkme slices (`EXTENSIONS` and `EXTENSION_MINI_TABLE_TO_DEF_INITS`) are gated
  off on Emscripten; `EXTENSION_MINI_TABLE_TO_DEF_INITS` elements are registered
  from `.init_array.101` constructors like `EXTENSIONS`, and `def_pool.rs`
  `load_all_extensions` reads those registrations on Emscripten.
- **linkme internals**: the Emscripten path relies on `#[doc(hidden)]
  linkme::DistributedSlice::private_new` and linkme_impl's `#![linkme_macro]` /
  `#![linkme_sort_key]` expansion grammar (verified against 0.3.35), so the OSS
  crate pins `linkme = "=0.3.35"` in
  `rust/release_crates/google_protobuf/Cargo-template.toml` (linkme itself pins
  `linkme-impl = "=0.3.35"` for the same reason). The exact pin applies to
  every OSS user, so a dependency graph that needs linkme >= 0.3.36 will not
  resolve until the Emscripten path is re-verified and the pin bumped; this is
  a deliberate trade-off.
- **Test coverage**: on TAP, `wasm_extension_test` runs natively (the
  `target_os = "emscripten"` code is compiled out) and `wasm_extension_build_test`
  only compiles and links `:wasm_extension_wasm`. Executing the Emscripten
  registration path (`node .../wasm_extension_bin.js`) and building
  `//third_party/protobuf/rust:protobuf` with upb on wasm32 are manual steps
  (see Tests Run).
- **Exported constructor globals on Emscripten**: On `wasm32-unknown-emscripten`,
  `#[used]` on `static __PROTOBUF_EXTENSION_CTOR` causes LLVM's Wasm backend to
  set `WASM_SYMBOL_EXPORTED`, so `wasm-ld`/`emcc` emits a Wasm global and export
  table entry per registered extension. Eliminate-at-compile-time retention
  without exporting would require `#[used(compiler)]` (unstable) or a `protoc`
  Rust generator change to emit one registration call per crate instead of per
  extension.

PiperOrigin-RevId: 1015838072