feat: add Integrated Lab page, app-header component, activation explainer, and UI improvements

feat: add Tiny-NLA activation explanation with trained model weights

init: TokenTrace - LLM interpretability toolbox