fix(tests): resolve 3 test failures in GUI decoupling track

- conftest.py: Create workspace dir before writing files (FileNotFoundError)
- test_live_gui_integration.py: Call handler directly since start_services mocked
- test_gui2_performance.py: Fix key mismatch (gui_2.py -> sloppy.py path lookup)

tests/test_gui2_performance.py

@@ -56,10 +56,12 @@ def test_performance_benchmarking(live_gui: tuple) -> None:
 
 def test_performance_baseline_check() -> None:
  """
- Verifies that we have performance metrics for gui_2.py.
+ Verifies that we have performance metrics for sloppy.py.
  """
- if "gui_2.py" not in _shared_metrics:
-  pytest.skip("Metrics for gui_2.py not yet collected.")
- gui2_m = _shared_metrics["gui_2.py"]
+ # Key is full path, find it by basename
+ gui_key = next((k for k in _shared_metrics if "sloppy.py" in k), None)
+ if not gui_key:
+  pytest.skip("Metrics for sloppy.py not yet collected.")
+ gui2_m = _shared_metrics[gui_key]
  assert gui2_m["avg_fps"] >= 30
  assert gui2_m["avg_ft"] <= 33.3