Add resilience controls for WeatherAPI/FCM dependency outages

[BadgerOps]

Problem

Weather and push paths depend on external services (WeatherAPI, FCM) with limited resilience controls, so transient outages can degrade UOTD recommendations and notifications.

Proposed Fix

• Add timeout + retry with jitter/backoff for external calls
• Add fallback behavior (last-known-good weather / safe default UOTD path)
• Add health/staleness alerts (e.g., weather cache expiration drift, scheduler failure rate)
• Document incident runbook for degraded mode

Acceptance Criteria

• Transient external failures do not hard-fail core workflows
• Stale weather cache and scheduler failures alert operators
• Degraded mode behavior is explicit and tested
• Runbook exists for on-call/admin recovery steps