Use full postgres image on minimal CNPG clusters so WAL archiving works

PR #298 added barmanObjectStore WAL archiving to every CNPG cluster, but
litellm, puppet and paperclip run the `17-minimal-trixie` image, which omits
the barman-cloud CLI. Their archiver dies with `exec: "barman-cloud-wal-archive":
executable file not found in $PATH`, so their buckets stay empty.

Switch those three to `17-system-trixie` (the -system variant already used by
the other clusters), which bundles the barman-cloud tools the in-tree backup
mechanism shells out to. This is a rolling image change; CNPG performs it as a
switchover with no data change.

Note: litellm and puppet will only start archiving once the separate stale
vault-ca-cert trust-anchor issue is also fixed (tracked separately) — that
blocks TLS to s3.ceph.unkin.net for all clusters.

Claude-Session: https://claude.ai/code/session_015ur3i7D2azsMAWTSVABApv

apps/base/litellm/cnpg_cluster.yaml

@@ -48,7 +48,7 @@ spec:
   enablePDB: true
   enableSuperuserAccess: false
   failoverDelay: 0
-  imageName: ghcr.io/cloudnative-pg/postgresql:17-minimal-trixie
+  imageName: ghcr.io/cloudnative-pg/postgresql:17-system-trixie
   instances: 3
   logLevel: info
   maxSyncReplicas: 0

apps/base/paperclip/cnpg_cluster.yaml

@@ -48,7 +48,7 @@ spec:
   enablePDB: true
   enableSuperuserAccess: false
   failoverDelay: 0
-  imageName: ghcr.io/cloudnative-pg/postgresql:17-minimal-trixie
+  imageName: ghcr.io/cloudnative-pg/postgresql:17-system-trixie
   instances: 3
   logLevel: info
   maxSyncReplicas: 0

apps/base/puppet/cnpg_cluster.yaml

@@ -69,7 +69,7 @@ spec:
   enablePDB: true
   enableSuperuserAccess: false
   failoverDelay: 0
-  imageName: ghcr.io/cloudnative-pg/postgresql:17-minimal-trixie
+  imageName: ghcr.io/cloudnative-pg/postgresql:17-system-trixie
   instances: 3
   logLevel: info
   maxSyncReplicas: 0