Record metric for "RESOURCE_EXHAUSTED" responses

[data-sync-user]

I noticed that we’re getting a number of RESOURCE_EXHAUSTED responses from FCM starting on or around 07-DEC-2024. These errors do not appear to be significantly impacting service, but they are unfamiliar and should be tracked. (They also appear to be from us sending too many messages per second to FCM, so it may be valid Back Pressure being applied. In which case, we may wish to treat these as back-off type errors.)

We should add a tracking metric for this sort of error to note if there’s any increase or decrease.

┆Issue is synchronized with this Jira Task