Add control-plane operator metrics and fix dashboard scraping

[albertompe]

Closes #56

Add control-plane operator metrics, fix metrics scraping, and revamp the operator dashboard

Summary

Introduces custom control-plane metrics for the Redkey operator, fixes a ServiceMonitor misconfiguration that left the operator dashboard empty, and reorganizes the operator Grafana dashboard for clarity.

New metrics

Five custom metrics are registered on the controller-runtime Prometheus registry (served on the existing operator /metrics endpoint) and instrumented at their reconcile call sites:

Metric	Type	Labels	Purpose
redkey_reconcile_stage_errors_total	counter	stage	Pinpoints which reconcile stage failed
redkey_config_creations_total	counter	reason	RedkeyConfig creations (initial vs generation_change)
redkey_cleanup_deleted_configs_total	counter	—	Configs removed by cleanup
redkey_robin_deployment_changes_total	counter	action	Robin Deployment create/patch operations
redkey_time_to_ready_seconds	histogram	—	Time from active config creation to Ready (observed on each transition into Ready)

All labels use bounded enums to keep cardinality predictable (no per-cluster/name labels).

Scraping fix

The operator ServiceMonitor scraped the metrics endpoint with scheme/port: https, but the operator serves plain HTTP on :8080 (--metrics-bind-address=:8080, service port named http). Prometheus therefore generated no target, so every panel showed "No data". Switched the ServiceMonitor to port/scheme: http to match, and updated the stale comment referencing the 8443/HTTPS endpoint.

Dashboard changes

• New Control-Plane Operations section (2 panels per row) grouping the existing operation panels plus the new metrics.
• Overview moved back to the top; all panel gridPos coordinates normalized to prevent section bleed-through (panels from one section rendering under another).
• Renamed "Superseded Configs Deleted" → "Configs Deleted by Cleanup" and clarified its description (it counts actual deletions, not configs merely in the Superseded phase).

Docs

Updated docs/observability.md with the new metrics, example queries, and the revised dashboard layout.

Testing

• go build ./..., go vet ./..., and go test ./internal/controller/... pass.
• Verified on a live cluster: operator target is up, and the new series (e.g. redkey_config_creations_total, redkey_phase) are ingested by Prometheus and rendered in Grafana.