Fix scheduler for cloud run

[amrit110]

This pull request updates the scheduler logic in backend/app/main.py to better support deployment on Cloud Run by decoupling background job execution from the application process. Instead, it introduces HTTP endpoints that can be triggered externally (e.g., by Cloud Scheduler), and ensures background jobs only run locally when explicitly enabled. It also improves shutdown handling for schedulers.

Key changes:

Cloud Run / Scheduler Adaptation:

• Background schedulers (ForecastScheduler and EvaluationScheduler) are no longer started by default on Cloud Run; instead, they are only enabled if the ENABLE_BACKGROUND_SCHEDULER environment variable is set to true. This prevents jobs from being killed when Cloud Run scales down.
• Informational log messages have been updated to clarify whether background schedulers are running or if HTTP endpoints should be used for triggering jobs.

New HTTP Trigger Endpoint:

• Added a new POST endpoint /scheduler/trigger-forecast to manually trigger a forecast run via HTTP. This endpoint is designed for use with Cloud Scheduler and ensures that forecast jobs run to completion within an active HTTP request. It includes logic to prevent duplicate runs if a job is already in progress.

Improved Scheduler Shutdown Handling:

• The shutdown logic now checks if schedulers are running before attempting to stop them, improving robustness when stopping the application.