Upload folder using huggingface_hub

Re-save model_q4.onnx with embedded weights (no external data, browser compat)

Add model_q4.onnx (renamed decoder_with_past_model for transformers.js v3 compat)