fix(daemon): bound CDP request waits

[honor2030]

Summary

• Bound regular daemon CDP request waits so a stuck Chrome/CDP call returns an error before the helper IPC socket times out.
• Add a regression test for a hanging Page.navigate path that previously left the helper with only a client-side _ipc.request() timeout.

Refs #370

Why

In #370, the visible symptom is a second navigation timing out in _ipc.py while the daemon later logs conn: Connection lost. I could not reproduce the underlying WebSocket drop on current main, but this hardens the daemon boundary so a stuck CDP send produces a useful Page.navigate timed out after 4.5s response instead of letting the helper hit its 5s socket deadline first.

Test Plan

• RED: uv run --with pytest --with pillow python -m pytest tests/unit/test_daemon.py::test_regular_cdp_calls_timeout_before_client_socket_deadline -q failed before the fix with an outer TimeoutError.
• GREEN: uv run --with pytest --with pillow python -m pytest tests/unit/test_daemon.py::test_regular_cdp_calls_timeout_before_client_socket_deadline -q → 1 passed.
• GREEN: uv run --with pytest --with pillow python -m pytest tests/unit -q → 91 passed.
• GREEN: uv run python -m compileall -q src tests → passed.
• GREEN: git diff --check → passed.
• Dynamic probe: BU_NAME=oss370pr2 uv run browser-harness with new_tab('https://example.com') followed by new_tab('https://httpbin.org/ip') → both navigations returned page_info() successfully on my local Chrome.

Notes

This is a scoped mitigation/diagnostic hardening PR, not a claim that the root WebSocket drop reported in #370 is fully fixed. If maintainers prefer a different timeout budget or want this limited to navigation/session-bound calls, I can adjust.

---

Summary by cubic

Bound CDP request waits in the daemon to 4.5s so stalled Chrome/CDP calls return a clear error before the helper’s 5s IPC socket deadline. Mitigates the timeout symptom in #370 by failing fast with messages like "Page.navigate timed out after 4.5s".

• Bug Fixes
  • Added CDP_CALL_TIMEOUT = 4.5 and a send_cdp() wrapper that uses asyncio.wait_for and returns a concise timeout error.
  • Used the wrapper for normal sends and after session re-attach.
  • Added a regression test ensuring the daemon surfaces the timeout before the client IPC socket deadline.

^{Written for commit 1689057. Summary will update on new commits. Review in cubic}

[cubic-dev-ai]

No issues found across 2 files

_{Re-trigger cubic}