MeetingSessionPOSTLogger failure handling needs work

[richnew10]

MSPL:

• Does not respect Retry-After headers, which are an important part of mitigating availability incidents.
• Never discards log messages on anything but a 200. Not only is this likely to be incorrect in the case of 4xx responses — it's a bad request, so for a misconfigured client or a poison pill it'll never recover — but it also trades memory to avoid data loss. Moreover this can make recovery harder after an availability incident — if the ingestion endpoint is unavailable for 10 minutes, it will receive all of that 10 minutes of backed-up traffic from clients as soon as it recovers.