Expand unit testing with hypothesis and edge case coverage

[monte-flora]

Add 46 new tests (109 total, up from 63) covering:

• ExplainToolkit init edge cases: unfitted models, bare estimators, numpy arrays, non-contiguous indices, y as list/Series
• PlotConfig API: set/get/reset, invalid keys, persistence
• Return type consistency: all compute methods return xarray.Dataset, computation_time_seconds in attrs
• Validation helpers: normalize_features/normalize_estimator_names with hypothesis property-based testing (random strings, lists)
• to_skexplain_importance: 1D/2D arrays, SHAP sum, normalization
• ALE property-based: varying n_features (2-8), n_bins (3-20), subsample fractions (0.1-1.0) via hypothesis
• Permutation importance: rankings length matches n_vars, uniqueness of rankings
• SAGE: basic, grouped, package availability check
• IO round-trip: save/load netCDF preserves values
• Grouping utilities: shape, scaling, local value grouping
• Degenerate inputs: constant features, single-feature datasets, 2-sample minimal datasets
• High-cardinality categorical ALE (issue Calculating ALE for high cardinality categorical feature throw a ValueError #86 regression tests)

Add hypothesis and sage-importance to requirements.txt for CI. Fix missing unittest import in test_interpret_curves.py.