feat(prometheus-direct-charting): slice 3 — remove grafana + config rewrite + changelog

Remove the entire Grafana surface: integrations/grafana.py, GrafanaWidgetSource
(+ _fetch_chart, now redundant since prometheus chart exists), GrafanaLinkWidget,
LinksTab, get_grafana_status endpoint, useGrafanaStatus hook, GrafanaStatus type,
fetchGrafanaStatus client fn, registry/nav/tab entries (FE+BE). Rewrite
config.yaml thin-dashboard rule to match reality (recharts is sanctioned for
Prometheus-backed series). CHANGELOG migration note added.

Backend: 293 pytest pass, ruff clean. Frontend: build+lint green (0 errors).
SC-115/116 grep-clean (only prometheus_range.py migration comments + Dashboard.test
shortcut fixture remain — both spec-allowed).

backend/src/media_library_viewer_api/services/settings_store.py

@@ -81,7 +81,7 @@ class SettingsStore:
             )
             conn.execute("CREATE INDEX IF NOT EXISTS idx_monitoring_machines_mode ON monitoring_machines(mode)")
             # The legacy SSH-scraping monitor (MonitoringPoller) was decommissioned;
-            # metrics now live in Prometheus/node_exporter/Grafana. Drop the orphan
+            # metrics now live in Prometheus/node_exporter. Drop the orphan
             # table on startup so existing databases get a clean slate.
             conn.execute("DROP TABLE IF EXISTS monitoring_machine_actions")
             conn.execute(