fix(cluster): 收敛宁波单实例数据库拓扑

将生产 CNPG 固定为宁波单实例，并显式关闭旧同步复制配置。发布流程在拓扑收敛前校验主库位置并创建 OSS 备份，应用后验证单实例、无复制连接和主库可用。同步更新生产监控、集群验收和发布脚本回归测试。

deploy/kubernetes/production/database.yaml

@@ -63,7 +63,9 @@ metadata:
     app.kubernetes.io/name: easyai-postgres
     app.kubernetes.io/part-of: easyai-ai-gateway
 spec:
-  instances: 2
+  # Production currently uses one Ningbo primary. Cross-region replicas are
+  # intentionally disabled until a separately certified HA topology exists.
+  instances: 1
   imageName: ghcr.io/cloudnative-pg/postgresql:18.4-standard-trixie@sha256:4587df73024408f5b2be9b4dd6ba2ccee8c9e5dc0c9a87c274c292291cc8a68c
   imagePullPolicy: IfNotPresent
   env:
@@ -106,6 +108,7 @@ spec:
     topologyKey: easyai.io/site
     nodeSelector:
       easyai.io/database: "true"
+      easyai.io/site: ningbo
   resources:
     requests:
       cpu: 500m
@@ -114,10 +117,9 @@ spec:
       cpu: "2"
       memory: 3Gi
   postgresql:
-    synchronous:
-      method: any
-      number: 1
-      dataDurability: preferred
+    # Keep this explicit null so server-side apply atomically removes the old
+    # two-instance synchronous setting during preflight field management.
+    synchronous: null
     parameters:
       # A 16 MiB WAL segment every 60 seconds requires at least 2.24 Mbit/s
       # of continuous cross-region throughput even while the database is idle.
@@ -129,10 +131,9 @@ spec:
       wal_compression: "on"
       wal_keep_size: 1GB
       # CNPG defaults these cross-region liveness timers to 5 seconds. A short
-      # control-plane or pod-network stall can otherwise tear down a healthy
-      # WAL stream and make preferred synchronous replication fall back to
-      # async. Keep failure detection bounded without amplifying a transient
-      # multi-second stall into repeated reconnects.
+      # control-plane or pod-network stall can otherwise tear down a future
+      # WAL stream. Keep failure detection bounded without amplifying a
+      # transient multi-second stall into repeated reconnects.
       wal_sender_timeout: 30s
       wal_receiver_timeout: 30s
   monitoring: