[FR] Add a typed error code to TopicManagementResponse.Error #1254

Is your feature request related to a problem? Please describe.
I subscribe tokens to topics and wanted to handle the token-related failures (mainly cleaning up stale tokens). This was harder than I expected, because TopicManagementResponse.Error only exposes a string (getReason()), and the possible values come from several different places:

• The values aren't documented. The Javadoc only says the reason is a non-null, non-empty string.
• Since 9.11.0 (feat(fcm): Migrate topic management to FCM v1 API #1240), subscribeToTopic() uses the FCM v1 API, while the *Legacy() methods still use Instance ID error codes. So a stale token can be reported as registration-token-not-registered (from NOT_FOUND) or as unregistered (from MessagingErrorCode.UNREGISTERED).
• The fallback uses toLowerCase() with the default locale, so with tr-TR, INVALID_ARGUMENT becomes ınvalıd-argument.

Describe the solution you'd like
A non-breaking addition:

• A new enum TopicManagementErrorCode (e.g. REGISTRATION_TOKEN_NOT_REGISTERED, INVALID_ARGUMENT, TOO_MANY_TOPICS, ..., UNKNOWN).
• TopicManagementResponse.Error#getErrorCode(), mapped from the raw server code so it doesn't depend on the locale. Unrecognized codes return UNKNOWN.
• getReason() and toString() stay as they are.

for (TopicManagementResponse.Error error : response.getErrors()) {
  switch (error.getErrorCode()) {
    case REGISTRATION_TOKEN_NOT_REGISTERED:
      removeToken(tokens.get(error.getIndex()));
      break;
    default:
      logger.warn("Topic subscription failed: {}", error.getReason());
  }
}

Open questions:

1. Should NOT_FOUND and UNREGISTERED both map to REGISTRATION_TOKEN_NOT_REGISTERED?
2. Should getErrorCode() return UNKNOWN or null (like FirebaseMessagingException#getMessagingErrorCode())?

Describe alternatives you've considered

• Only fixing the locale issue with toLowerCase(Locale.ROOT), but that's still string-based.
• Exposing a FirebaseMessagingException per error like SendResponse#getException(), but the legacy path has no per-token exception.

Additional context
Related: #1072, #1073. I have a patch with unit tests ready (existing tests unchanged, mvn clean test passes on JDK 8 and 17). Happy to open a PR if this direction sounds okay.