Scale down non-critical replicas for homelab resource savings

Reduce replicas to 1 for workloads that don't need HA in a homelab:
- Prometheus 2→1, Alertmanager 2→1 (~4.4GB RAM saved)
- cert-manager 3→1
- GitLab: registry 2→1, kas 2→1, sidekiq 2→1, praefect 2→1,
  pgbouncer-ro 3→1
- Guacamole + guacd 2→1
- Kiali 2→1, ArgoCD server 2→1
- Kyverno background-controller 2→1
- Scylla operator 2→1
- ext-authz-proxy 2→1, netbird-cluster-router 2→1

Kept multi-replica: coredns, envoy-gateway, kyverno admission,
vault, argocd-repo-server, gitlab-webservice, istiod, rook-ceph CSI.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

infrastructure/prometheus/values.yaml

@@ -3,8 +3,7 @@
 # Prometheus Operator configuration
 prometheus:
   prometheusSpec:
-    # High availability with 2 replicas
-    replicas: 2
+    replicas: 1
     
     # Reduce retention to lower I/O on shared NVMe storage
     retention: 7d
@@ -74,7 +73,7 @@ prometheus:
 alertmanager:
   enabled: true
   alertmanagerSpec:
-    replicas: 2
+    replicas: 1
 
 # Disable node-exporter due to PodSecurity restrictions
 prometheus-node-exporter: