Add README with model description, usage, and config reference

Add full model weights: vision + text encoder (fp16 and fp32)

Add GGUF models with LFS tracking

initial commit