Workers that fail unexpectedly leave `worker_runs` stuck in "running" state, causing phantom progress tracking

[lordofwar81]

When a worker fails unexpectedly (API error, timeout, kill), the worker_runs record stays in "running" forever. The tracking system shows phantom progress for dead workers.

Reproduction:

1. Spawn a worker that fails (API error, network issue)
2. Worker dies
3. Check DB: SELECT * FROM worker_runs WHERE status='running'; — record stuck
4. Status never updates to "failed"

Expected: Wrap worker execution in error handling that marks DB record as "failed" on any exception.

Workaround:

UPDATE worker_runs SET status='failed', completed_at=datetime('now') WHERE status='running' AND started_at < datetime('now', '-10 minutes');