feat(server): 可观测性 Phase 1 — 健康检查路由 + Prometheus 指标 + 连接池/事件积压监控

- 添加 /health/live 存活探针别名（原 /health + /health/ready 保留）
- 新增 metrics middleware：http_requests_total 计数器 + http_request_duration_seconds 直方图
- Prometheus exporter 独立端口 9090（可通过 ERP__SERVER__METRICS_PORT 覆盖）
- 后台任务每 30s 采样 DB 连接池活跃/空闲连接数（pg_stat_activity）
- 后台任务每 30s 采样 EventBus pending 事件积压数
- UUID 路径归一化避免高基数（/api/v1/users/:id/posts）

crates/erp-server/src/main.rs

@@ -432,6 +432,9 @@ async fn main() -> anyhow::Result<()> {
     // Start event cleanup (archive old published events + purge processed_events)
     tasks::start_event_cleanup(db.clone());
 
+    // Start DB connection pool metrics sampling (every 30s)
+    tasks::start_pool_metrics(db.clone());
+
     // Start timeout checker (scan overdue tasks every 60s)
     erp_workflow::WorkflowModule::start_timeout_checker(db.clone(), event_bus.clone());
     tracing::info!("Timeout checker started");
@@ -611,8 +614,13 @@ async fn main() -> anyhow::Result<()> {
     let app = Router::new()
         .nest("/api/v1", unthrottled_routes.merge(public_routes).merge(protected_routes))
         .nest("/uploads", uploads_router)
+        .layer(axum::middleware::from_fn(middleware::metrics::metrics_middleware))
         .layer(cors);
 
+    // Start Prometheus metrics exporter on a separate port
+    let metrics_port = state.config.server.metrics_port;
+    middleware::metrics::start_metrics_server(metrics_port);
+
     let addr = format!("{}:{}", host, port);
     let listener = tokio::net::TcpListener::bind(&addr).await?;
     tracing::info!(addr = %addr, "Server listening");