chore: reduce memory-db cluster from 3 to 2 instances

CHANGES:
- k8s/infra/databases/memory-db.yaml: instances 3 → 2
- Updated comment from '3 instances' to '2 instances'

REASONING:
- Reduces resource overhead (high availability at 2 is sufficient)
- Maintains quorum for failover (minimum 2 for HA)
- Saves memory/CPU allocation on homelab cluster
- ArgoCD will manage rollout automatically

DEPLOYMENT:
- ArgoCD will detect spec change and reconcile
- CNPG will scale down one pod
- Data preserved (3→2 replication, no data loss)

k8s/infra/databases/memory-db.yaml

@@ -1,6 +1,6 @@
 ---
 # CNPG Postgres cluster for Poimen Memory system (GitOps, declarative extensions).
-# 3 instances, pgvector 0.7.0 via spec.extensions (not manual CREATE EXTENSION).
+# 2 instances, pgvector 0.7.0 via spec.extensions (not manual CREATE EXTENSION).
 # Storage: 10Gi longhorn, consistent with temporal-db.yaml.
 # No manual psql needed — all via git/ArgoCD.
 apiVersion: postgresql.cnpg.io/v1
@@ -11,7 +11,7 @@ metadata:
   annotations:
     argocd.argoproj.io/sync-options: SkipDryRunOnMissingResource=true
 spec:
-  instances: 3
+  instances: 2
   imageName: ghcr.io/cloudnative-pg/postgresql:16.2
   enableSuperuserAccess: false
   storage: