fix: use parametrized test values in PrototypeRepresentationLearner tests

[BALOGUN-DAVID]

[Improvement] – Fix parameterized unit tests and add multi-dimensional sensitive features coverage for PrototypeRepresentationLearner

Description

Context:
The existing unit tests for PrototypeRepresentationLearner in Fairlearn had a critical gap: while @pytest.mark.parametrize decorators were defined to test multiple input variations (including sensitive_features=None and pd.Series labels), these parameterized values were immediately overwritten inside the test functions. As a result, key code paths—like the None sensitive features and string-based label handling—were never tested. Additionally, no tests existed for multi-dimensional sensitive features (e.g., race + gender), which are essential for intersectional fairness assessment in real-world scenarios.

Approach:

• Removed hardcoded overwrites in test_reconstruction and test_classification so pytest parametrize runs all intended cases.
• Added two new test functions:
  1. test_statistical_parity_multi_dimension_sensitive_features
  2. test_transform_sensitive_features_with_two_dimensions
     These validate behavior with 2D sensitive features.
• Simplified parametrize syntax for readability.
• Imported _merge_columns utility to correctly merge multi-dimensional columns, mirroring internal learner behavior.

Impact:

• Algorithmic Fairness: Properly tests both sensitive_features=None and multi-dimensional protected attributes.
• Data Integrity: Ensures string-based labels (pd.Series) pass through the pipeline correctly.
• Model Robustness: Multi-dimensional sensitive features now covered, guaranteeing the learner handles real-world intersectional groups.
• Test Coverage: Increased from 2 to 6+ parametrize combinations, plus 2 new tests.

---

Visual Proof / Evidence

Screenshots/GIFs should show pytest results with all parametrize combinations passing and coverage report highlighting new multi-dimensional sensitive features tests.

---

Tests

• New tests added
• Existing tests adjusted (removed overwrites)

Unit Tests Summary:

• test_reconstruction fixed
• test_classification fixed
• Added test_statistical_parity_multi_dimension_sensitive_features
• Added test_transform_sensitive_features_with_two_dimensions

---

---

[taharallouche]

LGTM thank you @BALOGUN-DAVID 🙏 Before merging, can you please add a short entry in the change log for this bug fix ?

[BALOGUN-DAVID]

Hi @taharallouche 👋 I've added the changelog entry, Please let me know if any changes are needed. Thanks!

[taharallouche]

@BALOGUN-DAVID the change in this PR is the bug fix for degenerate sensitive feature values, not test parametrization though. Could you please correct the change log ?

[BALOGUN-DAVID]

Done! I've updated the changelog entry. Thanks for the correction @taharallouche