Add GGUF quantized models (BF16, Q8_0, Q4_K_M) and update README with GGUF usage section

Upload tokenizer.json with huggingface_hub

initial commit