Fix Frigate Prometheus metrics & rebuild Grafana dashboard (#252)
## Summary
- **Prometheus scrape target:** Changed from `frigate.frigate.svc.cluster.local:5000` (broken after ringtail migration) to `nvr.ops.eblu.me` via HTTPS through Caddy on indri
- **Grafana dashboard:** Rebuilt for Frigate 0.17 metrics — 12 panels total:
- Row 1 (stats): Uptime, Inference Speed, Camera FPS, Detection FPS, GPU Usage, GPU Temp
- Row 2 (timeseries): CPU Usage, Memory Usage
- Row 3 (timeseries): Camera FPS + Skipped FPS, GPU Usage + Memory over time
- Row 4 (timeseries): Storage Usage, Detection Events (rate by camera/label)
## Deployment and Testing
1. Sync prometheus app on branch:
```
argocd app set prometheus --revision fix/frigate-metrics-dashboard && argocd app sync prometheus
```
2. Check `prometheus.ops.eblu.me/targets` — frigate job should show UP
3. Sync grafana-config:
```
argocd app sync grafana-config
```
4. Check `grafana.ops.eblu.me` — Frigate NVR dashboard should show live data
5. After merge: reset both apps to `--revision main` and sync
Reviewed-on: https://forge.ops.eblu.me/eblume/blumeops/pulls/252
Erich Blume
2c6c6a244a