Add comprehensive test suite for SEFR and LinearBoostClassifier estimators

[msamsami]

This PR introduces a comprehensive test suite for both SEFR and LinearBoostClassifier estimators and a few minor changes to the source code.

Comprehensive Test Suite

This test suite ensures that both estimators are robust, reliable, and fully compatible with the scikit-learn ecosystem while maintaining high code quality standards. The added test categories include:

• Initialization & Parameter Validation: Ensures proper setup with all parameter combinations
• Fitting & Training: Validates model training across different data scenarios
• Prediction & Inference: Comprehensive testing of all prediction methods
• Error Handling: Robust validation of error cases and edge conditions
• Mathematical Properties: Ensures numerical stability and algorithmic correctness
• Integration Testing: Dataset validation and sklearn compatibility

Minor Changes

• Removed unsupported class weight option: Dropped balanced_subsample from LinearBoostClassifier as it's not supported by Scikit-learn's compute_sample_weight.
• Added test coverage pragma: Added pragma comment to typing import in sefr.py and linear_boost.py.