Build a job queue with workers and retries

• enqueue(jobName, payload, { delayMs?, priority?, maxAttempts? }) adds a job and returns its id; jobs have states waiting, delayed, active, completed, failed and dead
• Workers register a handler with a concurrency limit; the queue runs at most that many jobs at once, and starts the next waiting job as soon as a slot frees up
• A job taken by a worker becomes invisible for a visibility timeout; if the worker does not acknowledge within it, the job becomes available again (the worker is presumed crashed), with the attempt count incremented
• When a handler throws or rejects, the job is retried with exponential backoff (delay grows per attempt, with a cap) until maxAttempts, then moved to a dead-letter queue with the last error
• ack(jobId) marks completion; a late acknowledgement from a worker whose visibility timeout already expired is rejected, so a job is never counted twice
• Introspection: stats() returns counts per state, getDeadLetters() lists dead jobs, and retryDead(jobId) re-enqueues one; time is driven by an injectable clock or scheduler for tests

Only after every requirement works.

• Priority queues and rate limiting per job type
• Repeatable (cron) jobs
• Graceful shutdown that waits for active jobs to finish