Add model card with Trac Network usage + original Qwen3.5 docs

Qwen3.5-4B 4-bit MLX quantized (mlx-vlm, multimodal VLM)

initial commit