fix: add scheduler daemon mode to handle missed cron jobs after system wake-from-sleep

[Jefsky]

Problem

When running TrendRadar in Docker with RUN_MODE=cron (the default), scheduled tasks rely on supercronic to fire python -m trendradar at the configured CRON_SCHEDULE times. This works fine as long as the host system stays awake.

If the host enters sleep/hibernation (laptop lid close, standby, etc.), supercronic does not fire jobs that were scheduled during the sleep period. When the system wakes up:

• The 11:00 scheduled run is simply skipped
• The user must manually docker restart the container
• After restart, the next scheduled time (e.g., 15:00) works again

This affects users on laptops (Mac/Windows/Linux) who use Docker Desktop or put their machines to sleep overnight.

Root Cause

supercronic (like all standard cron implementations) only fires jobs at their exact scheduled moment. There is no catch-up mechanism for missed jobs — the scheduling event is ephemeral.

Fix

Add a new RUN_MODE=daemon mode that replaces supercronic with a lightweight persistent Python scheduler daemon (docker/scheduler_daemon.py).

The daemon:

1. Parses the standard CRON_SCHEDULE expression (e.g., 0 11,15 * * *)
2. Runs a main loop that checks the time every 60 seconds
3. Fires TrendRadar at the scheduled times with deduplication
4. On startup, checks if any scheduled runs were missed within the last 2 hours and runs catch-up immediately — this handles wake-from-sleep
5. During runtime, continuously monitors for missed runs and catches up
6. Runs in parallel with the existing Web server (both start in the entrypoint)

How to use

Set the environment variable in your docker-compose.yml or Docker run command:

environment:
  RUN_MODE: daemon          # replaces the default "cron" mode
  CRON_SCHEDULE: "0 11,15 * * *"
  IMMEDIATE_RUN: "true"     # optional: run once on startup too

Testing

1. Deploy with RUN_MODE=daemon and CRON_SCHEDULE=0 11,15 * * *
2. Let it run and confirm scheduled execution at 11:00 and 15:00
3. Put the host system to sleep 10 minutes before a scheduled time
4. Wake the system after the scheduled time has passed
5. Observe that TrendRadar runs the catch-up immediately on wake
6. Check container logs for "[daemon] System wake-from-sleep detected: N missed run(s)"

Fixes #893