* math prompt improvements
* ignore brackets in complex_arithmetic results
* improve additional instruction in prompt of polynomial_equations
* more strict tests for score_answer in polynomial_equations
* simplify special reward handling
* fix test_intermediate_integration
* fix sokoban dataset
* add common dataset score_answer consistency test
* Add score_answer method to word_ladder
* add unit test for WordLadderDataset::score_answer()
---------
Co-authored-by: Andreas Koepf <andreas.koepf@provisio.com>
- Enhanced configuration validation with size and length constraints
- Implemented graph-based neighbor computation and caching
- Simplified path finding algorithm with more robust length checking
- Added more flexible word set loading with configurable length ranges
- Improved error handling for dataset generation
- Precomputed sorted word lists for each word length (stored in self.words_lists) to avoid redundant sorting on every _generate_word_pair call.
- Updated _generate_word_pair to utilize the cached sorted list, significantly improving computational efficiency.
- Implemented weighted random sampling for 5-letter words, giving words ending with 'S' a lower weight (0.5) to reduce bias without completely filtering them out.
