Qwen2-VL-7B on MLX — notes from a weekend of fighting bbox scaling

Ported Qwen2-VL-7B. Quick TL;DR:

• weights: q4 fits in ~5.2GB.
• image preprocessing: Qwen expects a very specific resize+crop.
• inference: ~14 tok/s on M2 Pro, ~26 tok/s on M3 Max.