feature request: database, cache and latency monitoring

[simonvanlierde]

Follow-up to #122. The central stack in CMLPlatform/monitoring covers traces, logs, host/container metrics and the core alerts. Still missing:

• Postgres and Redis metrics (connection pool, cache hit/miss), collected by Alloy next to the stack, plus a Database & cache dashboard
• Alert on p99 latency > 2s over 5 min
• Alert on sustained host memory > 90%

Alert rules and dashboards belong in the monitoring repo; exporter wiring belongs in deploy/alloy/config.alloy here.

[simonvanlierde]

CMLPlatform/monitoring#36 adds the p99 latency and host memory alerts, plus a DB connection pool panel built from the SQLAlchemy metrics the backend already exports.

Still open: Postgres server-side metrics and Redis hit/miss. Both need exporter containers with their own credentials (a pg_monitor role, the Redis password) and an optional scrape hook in the shared Alloy template.