Add Prometheus metrics endpoint and Grafana dashboard

[Copilot]

Adds observability to Gee Bee via a Prometheus metrics endpoint and a pre-built, auto-provisioned Grafana dashboard.

Metrics (internal/metrics/metrics.go)

Four counters registered via promauto and served at http://0.0.0.0:<port>/metrics:

Metric	Labels	Description
geebee_aircraft_spotted_total	tail_number	Aircraft spotted per registration
geebee_api_calls_total	—	Upstream ADS-B API calls
geebee_api_errors_total	—	Upstream ADS-B API errors
geebee_notifications_sent_total	channel	Notifications sent (discord, slack, custom_url, terminal)

Port defaults to 9090, configurable via METRICS_PORT env var or metricsPort in config.

Grafana Dashboard (grafana/provisioning/dashboards/gee-bee.json)

7-panel dashboard covering rate timeseries for aircraft spotted and notifications by channel, plus stat panels for all four counters. Datasource is templated so it works across Grafana instances.

Docker Compose

docker compose up brings up all three services — gee-bee (:9090), Prometheus (:9091), and Grafana (:3000) — with the dashboard and datasource auto-provisioned.

prometheus:
  image: prom/prometheus:latest
  volumes:
    - ./prometheus.yml:/etc/prometheus/prometheus.yml:ro

grafana:
  image: grafana/grafana:latest
  volumes:
    - ./grafana/provisioning:/etc/grafana/provisioning:ro

[rorpage]

Closing as no longer compatible with the project direction. The July 3 refactor ("Convert gee-bee to a one-shot binary scheduled by cron/systemd") moved gee-bee to a one-shot binary with no internal loop, and explicitly dropped Docker ("No Docker... Docker adds path, permission, and user complexity that the Go binary approach eliminates entirely" — see AGENTS.md). This PR adds a long-running :9090 metrics HTTP server and a Docker Compose stack (Prometheus + Grafana), both of which conflict with that architecture, and it now has git conflicts against main in Dockerfile (deleted upstream), README.md, and cmd/geebee/main.go. If metrics are wanted later, they'd need to fit the one-shot/no-Docker model (e.g. a Pushgateway or per-run stdout metric rather than a persistent scrape endpoint).

---

Generated by Claude Code