fix(gateway): suppress lifecycle status emails on email platform

[ibrahimokdadov]

Summary

• The gateway's _status_callback_sync sends every lifecycle event (rate limit retries, fallback attempts, error messages) through adapter.send()
• On chat platforms (Discord, Telegram, Slack) these are quick inline messages that get edited or disappear
• On email, each status update becomes a separate email, flooding the user's inbox with internal retry noise like:
  • ⏱️ Rate limit reached. Waiting 2.73s before retry (attempt 2/3)...
  • ⚠️ Max retries (3) exhausted — trying fallback...
  • ❌ Rate limited after 3 retries — HTTP 429: ...
• Fix: skip lifecycle events for the email adapter and log them instead

Change

gateway/run.py — _status_callback_sync():

if source.platform == Platform.EMAIL and event_type == "lifecycle":
    logger.info("[Email] status (%s): %s", event_type, message)
    return

Test plan

• Trigger a rate limit error while using the email gateway → verify no status emails are sent
• Verify the lifecycle events still appear in the gateway log
• Verify status messages still work normally on Discord/Telegram/other platforms

[teknium1]

Thanks for identifying the inbox-noise problem. The exact Email lifecycle rule is still not present on current main, although current main now selectively suppresses retry/fallback noise through gateway/run.py:440 and gateway/run.py:17812.

Problems

• gateway/platforms/email.py:363 marks a message \\Seen before the fetch at line 365. A failed fetch is then skipped after restart because this PR initializes its seen set from server-side SEEN messages. Mark it seen only after successful retrieval and handoff.
• Please add a regression test for the Email lifecycle behavior. Current main's shared filter suppresses only selected noisy strings; run_agent.py:875 still emits all statuses as lifecycle.
• The PR includes two unrelated IMAP behavior changes alongside the lifecycle fix. They should be split for independent review.

Suggested changes

• Port the lifecycle policy to current main's centralized callback/filter path (gateway/run.py:17812) and test it with Platform.EMAIL.
• Move the server-side \\Seen update after successful processing.

Automated hermes-sweeper review.

[teknium1]

This marks the UID \\Seen before the fetch at line 365. If that fetch fails, the message is skipped after restart because connect() now seeds _seen_uids from all server-side SEEN messages. Mark it seen only after successful fetch and handoff.