chore(cnpg): scale shared-cluster from 3 instances to 2

This sandbox sees low-demand workloads. 2 instances (1 primary + 1
streaming replica) still gives automated failover; the third replica is
unnecessary overhead — saves a pod, ~600Mi RAM, and a 20Gi block storage
volume.

When scaled down, cnpg will gracefully drain and remove
`shared-cluster-3` (the highest-numbered replica). The `shared-cluster-3`
PVC is retained per the `linode-block-storage-retain` storage class, in
case we ever need to recover the data; can be deleted manually if not
needed.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: themightychris

infra/cloudnative-pg/shared-cluster.yaml

@@ -4,7 +4,7 @@ metadata:
   name: shared-cluster
   namespace: cloudnative-pg
 spec:
-  instances: 3
+  instances: 2
   imageName: ghcr.io/cloudnative-pg/postgis:18-3-system-trixie
   storage:
     size: 20Gi