hermes - 💡(How to fix) Fix Cron accepts one-shot jobs in the past and leaves them scheduled with no next_run_at

Summary

A one-shot cron job can be created with a run_at timestamp that is already in the past. Hermes accepts it, but the job becomes a zombie:

• enabled: true
• state: scheduled
• repeat: once / completed: 0
• next_run_at: null
• last_run_at: null
• last_status: null

It never fires and shows up later as a stuck scheduled job.

Repro / observed example

On a live Hermes gateway, a reminder job was present with:

{
  "id": "c8d893b0a214",
  "name": "recordatorio-macbook-sequoia-manana-9",
  "schedule_display": "once at 2026-05-10 09:00",
  "repeat": { "times": 1, "completed": 0 },
  "enabled": true,
  "state": "scheduled",
  "created_at": "2026-05-12T19:51:43.338068+02:00",
  "next_run_at": null,
  "last_run_at": null,
  "last_status": null
}

The job was created after its scheduled time, so it had no valid future execution.

Expected behavior

One of these would be fine:

1. Reject one-shot schedules in the past at creation/update time with a clear error.
2. Mark them immediately as expired/missed/disabled instead of scheduled.
3. If intentional, expose an explicit missed status so health checks do not treat it as a live scheduled job.

Why it matters

The current state is misleading: it looks scheduled/enabled but can never run. Health checks/heartbeat jobs then report it as stuck, and users have to manually remove it.