Call va_end even if OPENSSL_vasprintf_internal fails

Though malloc failures are practically impossible and I believe va_end
is a no-op on all platforms we support, it is technically UB to call
va_start without calling va_end. Fix that.

Change-Id: I326f7d3c5c7dab3e6200c4ce2c483c1dd91778ae
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/102707
Auto-Submit: David Benjamin <davidben@google.com>
Reviewed-by: Adam Langley <agl@google.com>
Presubmit-BoringSSL-Verified: boringssl-scoped@luci-project-accounts.iam.gserviceaccount.com <boringssl-scoped@luci-project-accounts.iam.gserviceaccount.com>
Commit-Queue: Adam Langley <agl@google.com>

crypto/err/err.cc

diff --git a/crypto/err/err.cc b/crypto/err/err.cc
index 997399e..fc30f01 100644
--- a/crypto/err/err.cc
+++ b/crypto/err/err.cc
@@ -670,10 +670,11 @@
   va_list ap;
 
   va_start(ap, format);
-  if (OPENSSL_vasprintf_internal(&buf, format, ap, /*system_malloc=*/1) == -1) {
+  int len = OPENSSL_vasprintf_internal(&buf, format, ap, /*system_malloc=*/1);
+  va_end(ap);
+  if (len == -1) {
     return;
   }
-  va_end(ap);
 
   err_set_error_data(buf);
 }