Safer Worker Shutdowns and Restarts

This release strengthens the Worker shutdown and restart lifecycle. worker.stop() now waits until the active processing session is fully quiescent before returning.

Fixed

• Drain Jobs acquired while shutdown is already in progress, ensuring no work escapes graceful shutdown.
• Wait for in-flight acquisitions, executions, finalizations, schedule dispatches, and heartbeat renewals before completing shutdown.
• Keep renewing leases for running or late-acquired Jobs until they finish.
• Serialize restarts with shutdown so a new session cannot acquire Jobs before the previous one has fully stopped.
• Prevent a Worker stopped during initialization from acquiring work.
• Interrupt idle and error delays immediately when stopping.
• Prevent additional Schedule claims and cycle events once shutdown begins.
• Observe Job execution failures as soon as they enter the pool, preventing transient unhandled promise rejections.

Changed

• Each processing run is now isolated in an internal Worker session with explicit lifecycle ownership.
• Heartbeat management has been extracted into a dedicated internal component.
• Queue selection and processing mode are now fixed for the duration of a session. Call worker.stop() before changing queues or switching between manual (processCycle) and continuous (start or process) processing.

No configuration changes or migrations are required.