v5.9.6: Enhance stuck job monitoring and harden job state transitions

Add `READY` and `WAITING` jobs to the stuck jobs report. This detects cases where jobs are repeatedly returned to the broker but never processed, providing better visibility into potential system stalls.

Implement deadlock retries for `WAITING` state transitions and, if all retries fail, republish the job to the broker. This prevents premature `PERMANENTLY_FAILED` status for jobs whose callbacks never started.

Make the `PERMANENTLY_FAILED` state transition in `logException()` conditional on the job still being `READY` in the database, preventing overwrites of states changed by other processes.