Release v0.9.0: Memory-Efficient Chunked Training System

- Add memory-efficient chunked training system for large datasets
- Prevent memory accumulation during training on 100K+ logs
- Automatic chunking with ~900KB chunks stored in temp files
- Bounded memory usage (~25-50MB) regardless of dataset size
- Update documentation and version badges
- Add comprehensive demo scripts for chunked training
- Fix code quality issues and improve error handling