[Feature Request] Scheduled-task run timeout + retry/backoff policy

[cagdasyurekli]

Problem

A scheduled run that hangs (e.g. an approval parked in the inbox) stays in running status forever. The scheduler\x27s overlap guard (_running_ids, coworker/automation/scheduler.py) is only released when the runner returns, so a hung run silently skips all future runs of that task. #228 fixed the specific always_allowed_commands case but explicitly left the general problem open: no run timeout and a non-resilient overlap guard.

Proposed Solution

1. Add a per-task timeout (default e.g. 15 minutes) after which a run is marked timed_out, releasing the overlap guard and parking a notification for the user.
2. Add optional max_retries with backoff for runs that end in error (not user-cancelled or timed-out).
3. Add an explicit force_stop action so a user can halt a stuck run from the UI.

Implementation Scope & Alignment

• Scope: ~150-200 lines in coworker/automation/scheduler.py + store.py + a small GUI control.
• Zero New Dependencies.
• Alignment: Keeps the "standing automations" promise honest — a stalled automation should not silently kill every future run of that task.