Model card: q4f16_1 -> q4f32_1

Remove stale q4f16 lib (replaced by q4f32_1)

Add q4f32_1 WebGPU lib

Rebuild as q4f32_1 (fixes f16-overflow NaN in get_last_hidden)

Upload recursivelink.json with huggingface_hub

Add WebGPU model library

Add model card

Upload folder using huggingface_hub

initial commit