perf(worker): 按实例内存上限扩展异步容量

将策略推导出的集群目标与单 Worker 内存安全上限分离，避免失活实例的全部容量转移到单个 2 GiB Pod。

生产环境单实例上限和媒体物化并发设为 24；两实例正常时总容量由 16 提升至 48，后续可通过增加 Worker 节点继续扩展。新增 0093 迁移保存实例容量上限，并兼容滚动发布中的旧实例。

风险：更高媒体并发会增加 Worker 和节点内存压力，保留单实例 24 的硬边界并由持续监控观察 RSS、OOM、队列与节点余量。

验证：go test ./... -count=1；go vet ./...；真实 PostgreSQL 分配与故障转移测试；迁移安全检查；kubectl kustomize；gofmt -l；git diff --check。

deploy/kubernetes/production/application-config.yaml

@@ -35,5 +35,5 @@ data:
   AI_GATEWAY_LOCAL_RESULT_MAX_BYTES: "268435456"
   AI_GATEWAY_LOCAL_RESULT_MAX_TASK_BYTES: "536870912"
   AI_GATEWAY_ASYNC_QUEUE_WORKER_ENABLED: "true"
-  AI_GATEWAY_ASYNC_WORKER_HARD_LIMIT: "16"
+  AI_GATEWAY_ASYNC_WORKER_HARD_LIMIT: "2048"
   AI_GATEWAY_ASYNC_WORKER_REFRESH_INTERVAL_SECONDS: "5"

deploy/kubernetes/production/application.yaml

@@ -312,8 +312,10 @@ spec:
               value: "true"
             - name: AI_GATEWAY_DATABASE_MAX_CONNS
               value: "24"
+            - name: AI_GATEWAY_ASYNC_WORKER_INSTANCE_HARD_LIMIT
+              value: "24"
             - name: AI_GATEWAY_MEDIA_MATERIALIZATION_CONCURRENCY
-              value: "8"
+              value: "24"
             - name: POD_NAMESPACE
               valueFrom:
                 fieldRef:
@@ -439,8 +441,10 @@ spec:
               value: "true"
             - name: AI_GATEWAY_DATABASE_MAX_CONNS
               value: "24"
+            - name: AI_GATEWAY_ASYNC_WORKER_INSTANCE_HARD_LIMIT
+              value: "24"
             - name: AI_GATEWAY_MEDIA_MATERIALIZATION_CONCURRENCY
-              value: "8"
+              value: "24"
             - name: POD_NAMESPACE
               valueFrom:
                 fieldRef: