Cover _fetch_local_models (LM Studio and Ollama parsing, error paths,
missing base_url), _fetch_lmstudio_available_models (happy path and
errors), and _load_lmstudio_model (success, API failure, exception).
All mocked via monkeypatch/MagicMock — no real HTTP calls.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>