Update documentation

Default ASL detection to WLASL auto backend

Add experimental WLASL I3D backend

Add sliding-window ASL phrase prototype

Expose rejected ASL candidates in intent

Speed up live ASL debug diagnostics

Use single 30-frame ASL upload inference

Draw MediaPipe landmarks in live debug

Align ASL inference with upstream model

Harden ASL confidence and LLM consistency

Integrate ASL TFLite model assets

Remove demo gloss prefill and harden no-detection flow

Expose stepwise ASL debug pipeline

Fix Gradio 6 video input handling

Refactor app into pipeline bricks

Import ASL video pipeline