Node SQLite Worker Pool

A durable job queue on node:sqlite feeding a worker_threads pool, where jobs survive process restarts: any row left running by a crash is re-queued on boot. The same file is its own worker through isMainThread, and the main thread claims a job with an atomic UPDATE ... RETURNING only when a worker is idle, so backpressure is structural rather than a buffer that grows without bound. Failures retry with capped exponential backoff plus full jitter, then land in a dead state, and SIGINT drains in-flight jobs before terminating workers and closing the DB. Ships a node:test suite that uses mock timers and a mocked clock to prove the backoff fires at exactly the computed delays and that a failed job stays unclaimable until its window elapses, with no real waiting.