Improve sensitivity coverage with focused executor unit tests

[Copilot]

Summary

Adds focused unit tests for py3plex/sensitivity/executor.py and metric-spec parsing to improve coverage in the low-covered sensitivity area.

Changes

• Added new test module: tests/test_sensitivity_executor.py
• Added coverage for:
  • _extract_conclusion_data (dict input, to_dict() object input, uncertainty-wrapped values)
  • _compute_stability_metric dispatch and unknown-metric fallback
  • _find_collapse_point found/not-found behavior
  • _compute_local_influence per-node score behavior
  • run_sensitivity_analysis deterministic flow under monkeypatched perturbation
  • run_sensitivity_analysis instability detection (non-perfect stability)
  • parse_metric_spec variants
• Kept tests lightweight and deterministic with monkeypatched perturbation behavior and direct sensitivity module loading in the test context.

Validation

• Ran targeted tests successfully:
  • python3 -m pytest tests/test_sensitivity_executor.py tests/property/test_sensitivity_properties.py -q

Notes

• In this sandbox environment, coverage-plugin runs hit an environment-specific NumPy import issue (cannot load module more than once per process) when measuring coverage directly, so validation used targeted pytest execution.