perf: 小优化组 + 恢复浏览器页签

1. MCP 空闲睡眠 50ms -> 10ms：#187 后单轮 drain 约 0.3ms，
   100 次/s 空闲迭代约 3% 单核，请求与远控输入拾取延迟
   从平均 25ms 降到 5ms。
2. stream 轮询 50ms -> 20ms。
3. stream payload 加 url/title；面板 stream 处理合并 url/
   title，5 秒 state 轮询只在未观看时跑（观看时 stream 已
   全量覆盖，轮询纯冗余；不观看时保留轮询，避免会话开始
   后页眉 stale）。
4. 恢复浏览器页签：#176（ccb5a86）改写 TaskDetailPanel
   页签时顺带删了 BrowserPanel 的挂载，导致线上控制台
   自 #176 起就没有浏览器页签（面板文件一直在，只是不
   被 import，构建产物因此不变）。本次把 import、页签
   与 Card 原样挂回；构建产物含“开始观看”标记可证。
console.html 由 npm run build:console 重打（342484 字节）。
console-redesign/ 为构建脚本审阅产物，不进仓库。

cloud/playwright_remote.py

@@ -34,7 +34,7 @@ Threading contract (unchanged from the socket version, measured
 2026-10-30: Playwright sync calls from any thread but the
 creating one raise "Cannot switch to a different thread"):
 only the MCP main thread touches the browser. It calls
-RemoteBridge.poll() on every idle main-loop iteration (~50ms),
+RemoteBridge.poll() on every idle main-loop iteration (~10ms),
 so control/input/frames flow even while the agent is idle, and
 CDP frame acks keep flowing (no ack = CDP stops after the
 first frame).
@@ -57,7 +57,7 @@ CONTROL_NAME = ".browser-remote-control.json"
 STATE_STALE_SECONDS = 30
 # Heartbeat interval for state writes. The server treats state
 # older than STATE_STALE_SECONDS as dead, so 0.5s is plenty;
-# writing every ~50ms iteration was pure gofer round-trips.
+# writing every iteration was pure gofer round-trips.
 STATE_HEARTBEAT_SECONDS = 0.5
 
 
@@ -190,7 +190,7 @@ class RemoteBridge:
         """One drain-loop iteration: control, inputs, frame, state.
 
         Writes are change-gated. The old code rewrote both files
-        on every ~50ms iteration: ~20 atomic temp+replace
+        on every iteration: ~200 atomic temp+replace
         round-trips per second through the runsc gofer mount for
         identical bytes (pump returns the latest frame
         persistently, and updated_at changes every call). Now the
@@ -269,12 +269,14 @@ def read_state(workdir):
 def effective_holder(workdir, state):
     """Server-side holder with the control race closed.
 
-    The MCP rewrites the state file every ~50ms, so right after
-    a takeover the state file still shows the old holder. The
+    The MCP polls every ~10ms idle and rewrites the state file
+    at once when the holder changes (a key change bypasses the
+    heartbeat throttle), so right after a takeover the state
+    file still shows the old holder for at most ~10ms. The
     control file is the newer truth when its mtime is at or
     after the state's updated_at; otherwise the state file
     (already refreshed by a poll) wins. Both agree once the
-    MCP has polled, so this only closes the ~50ms window.
+    MCP has polled, so this only closes the ~10ms window.
     """
     try:
         mtime = os.path.getmtime(os.path.join(workdir, CONTROL_NAME))