feat(alertd): add database health check with database-down event

[passcod]

Summary

Adds a periodic health check that monitors the PostgreSQL database connection and triggers an internal database-down event when the database becomes unreachable.

Changes

New database-down event type

• events.rs — DatabaseDown variant on EventType (serialized as "database-down") and EventContext (carries database_url and error_message). Includes a default template mentioning the redacted DB URL, error details, and that all SQL-based alerts are non-functional.

Periodic health check in the daemon

• daemon.rs — Spawns a task that runs every 30 seconds:
  • Attempts pool.get_timeout(5s) then SELECT 1
  • Tracks was_down state so the event only fires on the healthy → unhealthy transition (no repeated alerts)
  • Logs when the database connection is restored
  • Redacts the password from the database URL before including it in the alert context

Re-exports

• lib.rs — Re-exports AlwaysSend, WhenChanged, TargetConnection, and TargetEmail for downstream/test use.

Tests

• tests/database_health.rs — 9 new tests:
  • Event type parsing and serialization
  • Alert definition with event: database-down
  • Default template rendering with expected content
  • Target resolution for database-down event alerts
  • Health check detection of unreachable databases
  • Health check success on valid databases
  • URL password redaction (with password, without password, unparseable)

Users can also define custom alert definitions for event: database-down in their alert YAML files to override the default template or route to specific targets.