Full training results & evaluation with BERTScore

Add evaluation_report.json for metrics display

Fixed compiling issue, added legnth penalty, and atttempting freezing encoder layers 0-5 to lower parameters and preserve T5's langauge understanding.

Update Gradio demo, inference factory, and evaluation results

Improve summarization output quality

Full training run, code cleanup, mypy/ruff fixes

Training run: dev config 1 epoch results

feat: Add FLAN-T5 compatibility with relative position bias

Refactor: Consolidate dependencies, improve testing, and add CI/CD