fix(tui_gateway): back off notification poller when session is busy

The busy-session branch of _notification_poller_loop re-queued the
completion event and immediately re-polled it with no sleep, spinning
at full speed (100% CPU, ~1100 futex/s of GIL churn) for as long as
the session stayed running. This starved the dashboard asyncio loop:
/api/status went from 0.14s to 3-6s with 10s timeouts.

Sleep 0.25s outside history_lock before re-polling, mirroring the
0.1s back-off already used for foreign-session events.

tui_gateway/server.py

@@ -8487,11 +8487,19 @@ def _notification_poller_loop(
             _emit("status.update", sid, {"kind": "process", "text": text})
             _emitted.add(_dedup_key)
 
+        _requeued = False
         with session["history_lock"]:
             if session.get("running"):
                 process_registry.completion_queue.put(evt)
-                continue
-            session["running"] = True
+                _requeued = True
+            else:
+                session["running"] = True
+        if _requeued:
+            # Back off before re-polling: the re-queued event keeps the queue
+            # non-empty, so without a sleep this loop spins at full speed
+            # (100% CPU, GIL churn) for as long as the session stays busy.
+            time.sleep(0.25)
+            continue
 
         rid = f"__notif__{int(time.time() * 1000)}"
         try: