Fix stale-task reaper to avoid duplicate publish requeues (
1624825)
The publish task reaper now updates task timestamps when it re-drives stale work, so the same pending jobs no longer get re-enqueued every few minutes once the queue falls behind. The fix also tightens the task-state handling around stale processing, which matters because the bug was causing duplicate work, failed retries, and queue congestion.
Other misc changes
- SQLx metadata regenerated for the updated publishing-task queries
- Minor Cargo.toml / Cargo.lock adjustments
- Test and database helper updates to match the reaper/publish logic changes