Default the allowed Prometheus host to kube-prometheus-stack's

v0.49.0 shipped the metrics feature with an empty allowlist, so nothing
showed on an incident until someone set prometheus.allowedHosts. The chart
now defaults it to kube-prometheus-stack-prometheus.kube-prometheus-stack:9090,
the Prometheus that stack installs. An empty list still turns the feature
off.

Chart default only: the server's own default (TERDUT_PROMETHEUS_ALLOWED_HOSTS
empty) is unchanged, and no Go code, API or schema changed.

Claude-Session: https://claude.ai/code/session_016mBLURvJoMuUEr9cB2RpUN

charts/terdut-server/values.yaml

@@ -120,8 +120,11 @@ operatorMode: false
 # empty turns the feature off. Entries are host names, with "*" as one DNS label
 # and a leading "**" as one or more (e.g. "*.monitoring.svc.cluster.local",
 # "prometheus-*.monitoring"), IPs or CIDR ranges, each with an optional ":port".
+# The default is the Prometheus that kube-prometheus-stack installs; set it to
+# [] to turn the feature off.
 prometheus:
-  allowedHosts: []
+  allowedHosts:
+    - kube-prometheus-stack-prometheus.kube-prometheus-stack:9090
   # How far before an alert started its graph begins.
   preroll: 1h
   timeout: 10s