fix: worker 进程并发 double-close panic + 状态页连接语义重构 + 集群页骨架

- process: spawn 不再重复起 supervise goroutine（Restart 触发旧 supervise close doneCh 后再 close → panic）；新增 closeOnce/supervising 守卫；Start 首次起 supervise，重启循环在同一 goroutine 内

- httpapi: status 增加 connState（connected/connecting/failed/not_started/disabled），由 worker 日志探测真实 frps 连接（不再把本地 worker 状态误当远程 frps 状态）

- StatusView: 远程卡片主状态改为连接状态徽标；启停按钮标注为控制本地 frpc worker；本地转发区保留 worker 状态显示

- 新增 ClusterView 集群页骨架（M3 LB+健康检查载体）+ App.vue 导航接入

- render: bandwidthLimit 从 proxy 顶层移入 transport 段（frpc 0.61 正确 schema，修复真实连不上的 bug）

internal/process/process.go

@@ -47,12 +47,14 @@ type worker struct {
 	proc    *exec.Cmd
 	logFile *rotatingFile
 
-	stopOnce sync.Once
-	stopCh   chan struct{}
-	doneCh   chan struct{}
+	stopOnce  sync.Once
+	closeOnce sync.Once
+	stopCh    chan struct{}
+	doneCh    chan struct{}
 
-	mu     sync.Mutex
-	status Status
+	mu          sync.Mutex
+	status      Status
+	supervising bool // guarded by Manager.mu
 }
 
 // Manager supervises all workers.
@@ -103,6 +105,12 @@ func (m *Manager) Start(name string) error {
 		m.mu.Unlock()
 		return err
 	}
+	m.mu.Lock()
+	if !w.supervising {
+		w.supervising = true
+		go m.supervise(w)
+	}
+	m.mu.Unlock()
 	return nil
 }
 
@@ -145,12 +153,13 @@ func (m *Manager) spawn(w *worker) error {
 	w.proc = cmd
 	w.logFile = logWriter
 	m.setState(w, Status{State: "running", Pid: cmd.Process.Pid, StartTime: time.Now().Unix(), RestartCount: restartCount})
-	go m.supervise(w)
+	// NOTE: no new supervise goroutine here. The first Start()'s supervise loop
+	// re-spawns on restart; spawning here would race on doneCh close.
 	return nil
 }
 
 func (m *Manager) supervise(w *worker) {
-	defer close(w.doneCh)
+	defer w.closeOnce.Do(func() { close(w.doneCh) })
 	for {
 		err := w.proc.Wait()
 		exitCode := 0