Handle Ordinis outage in TGU service

Classify Ordinis classifier transport failures and empty responses as unavailable, keep stale platform cache when possible, and expose Prometheus metrics/alerting for the outage signal.

helm/pcp-tgu-service/templates/prometheusrule.yaml

@@ -0,0 +1,21 @@
+{{- if .Values.prometheusRule.enabled }}
+apiVersion: monitoring.coreos.com/v1
+kind: PrometheusRule
+metadata:
+  name: {{ include "pcp-tgu-service.fullname" . }}-alerts
+  labels:
+    {{- include "pcp-tgu-service.labels" . | nindent 4 }}
+spec:
+  groups:
+    - name: {{ include "pcp-tgu-service.name" . }}.ordinis
+      rules:
+        - alert: PcpTguOrdinisUnavailable
+          expr: pcp_tgu_ordinis_unavailable == 1
+          for: {{ index .Values.prometheusRule.ordinisUnavailable "for" | quote }}
+          labels:
+            severity: {{ .Values.prometheusRule.ordinisUnavailable.severity | quote }}
+            service: {{ include "pcp-tgu-service.name" . | quote }}
+          annotations:
+            summary: "Ordinis classifier is unavailable for pcp-tgu-service"
+            description: "The last pcp-tgu-service attempt to fetch platforms from Ordinis failed. Existing stale platform cache is kept when available."
+{{- end }}

helm/pcp-tgu-service/values.yaml

@@ -35,5 +35,11 @@ resources: {}
 podAnnotations: {}
 podLabels: {}
 
+prometheusRule:
+  enabled: false
+  ordinisUnavailable:
+    for: 5m
+    severity: warning
+
 nameOverride: ""
 fullnameOverride: ""