fix(worker): 增强队列启动与租约到期恢复

River 初始化遇到瞬时数据库超时时最多重试三次，并确保启动异常时先取消后台 LISTEN 和 Worker 上下文再关闭连接池，避免不就绪 Pod 卡在退出路径。Worker 维护循环每 15 秒执行幂等运行时恢复，使旧实例中断任务在执行租约到期后自动重排并继续触发孤儿 Job 救援。验证通过完整 Go 测试、go vet、竞态检查和迁移安全检查。

apps/api/cmd/gateway/main.go

@@ -36,14 +36,17 @@ func main() {
 	}
 
 	ctx, stop := signal.NotifyContext(context.Background(), syscall.SIGINT, syscall.SIGTERM)
-	defer stop()
 
 	db, err := store.ConnectWithMaxConns(ctx, cfg.DatabaseURL, cfg.DatabaseMaxConns)
 	if err != nil {
+		stop()
 		logger.Error("connect postgres failed", "error", err)
 		os.Exit(1)
 	}
 	defer db.Close()
+	// Cancel background LISTEN and worker goroutines before closing the pool.
+	// This also prevents a startup panic from waiting forever in pgxpool.Close.
+	defer stop()
 	if recovery, err := db.RecoverInterruptedRuntimeState(ctx); err != nil {
 		logger.Error("recover interrupted runtime state failed", "error", err)
 		os.Exit(1)