refactor: delete dead sync worker + entrypoints (keep live sync/dispatch.py)

[jlunder00]

Summary

Scoped-down version of brief 0d. The original brief's pre-check ("git grep clean outside sync/") was wrong: sync/dispatch.py is actively imported and called by api/routes/integrations.py (dispatch_sync, manual-sync trigger + initial_sync_and_register) and has live test coverage in tests/api/test_integrations.py. Deleting it would have been a real behavior regression, not a cleanup — confirmed with team lead, scoping to the actually-dead pieces.

Deleted (the vestigial LISTEN-loop worker, confirmed absent from prod per its own docstring — "Fly.io/production deployments no longer require this process... [program:sync] section intentionally absent from production config", superseded by inline FastAPI BackgroundTasks since Phase F):

• sync/worker.py (219 lines — SyncWorker, PG LISTEN loop, APScheduler cron jobs)
• sync/main.py + sync/__main__.py (the python -m sync entrypoint)
• tests/sync/test_worker.py + tests/sync/__init__.py (now-empty test package)

Kept untouched:

• sync/dispatch.py — live dispatch mechanism (PG NOTIFY), imported by api/routes/integrations.py
• api/routes/integrations.py — out of scope for this cleanup, not touched
• Migration d2e3f4a5b6c7_sync_worker_integration_lookup.py (DB artifact, per brief)

No doc/config references to the deleted worker were found (grepped .md, supervisord/fly config) — nothing to update.

Flagged for follow-up (not investigated, out of scope here): since the LISTEN worker isn't running in prod, dispatch_sync's PG NOTIFY may have no consumer there — the manual-sync/initial-sync paths may be effective no-ops in prod, with real sync handled by the inline BackgroundTasks path since Phase F. Worth a dedicated investigation, not a cleanup-scope fix.

Test plan

• Full pytest suite: 12 pre-existing failures, identical to origin/dev baseline (missing claude_agent_sdk module + in-flight SDK-migration test failures unrelated to this change) — zero new failures.
• grep confirms no remaining references to sync.worker/sync.main/SyncWorker anywhere in the codebase.

[jlunder00]

Addressed code review findings in commit c6f6974:

1. docker-compose.yml sync: service — deleted. It ran python -m sync with restart: unless-stopped and no profiles guard; would have crash-looped on any docker compose up post-deletion.
2. systemd/tether-sync.service — deleted. Confirmed dormant (not one of the three deployed services per CLAUDE.md: tether-bot/tether-api/tether-mcp; no active CI references it).
3. Lost test coverage for the kept sync/dispatch.py — relocated the two direct dispatch tests (test_dispatch_sync_issues_pg_notify, test_dispatch_sync_webhook_type) from the deleted tests/sync/test_worker.py into new tests/sync/test_dispatch.py. These were the only tests exercising dispatch.py's real body (everything in tests/api/test_integrations.py monkeypatches it away).

Also fixed a stale sync/__init__.py docstring that still described the removed LISTEN worker.

Full suite re-verified: same 12 pre-existing baseline failures (unrelated to this change), 823 passed (821 + 2 relocated dispatch tests), zero new failures.