Medium training run

Major improvements: real titles, anti-overfitting, better demo UX

Add English language filter to all data downloads

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

Clean up codebase and fix training bugs

Update LexiMind: improved training, model architecture, and evaluation

Expand datasets: GoEmotions (28 emotions), Yahoo Answers (10 topics), BookSum

Full training run, code cleanup, mypy/ruff fixes

Style: Apply ruff formatting

Style: Fix linting errors and organize imports (ruff & mypy)

Summarization fix through pipeline

chore: snapshot current refinements