v1.0.11

Fixes for production stuck-task and queued_requests leak.

• worker_loop: persist started_at on the Task object and write status='processing' to task:{id} so _store_final_task_state keeps them. Previously started_at was lost in the final stored record.
• requeue_stuck_processing_tasks: fall back to queued_at when started_at is missing. Covers workers killed in the SADD processing_tasks → SET task:{id} window — those tasks were unrecoverable before.
• _store_final_task_state: ZREM queued_requests on terminal state. Stops the unbounded zset growth (some clusters had reached >1M entries).