Add feature

[shivdbz2010]

joycaption beta and qwen3 vl
remove background - fill transparent color with fixed/random color , gaussian blur background
crop / mask include exclude / label overlay via SAM/Yolo seg,bbox detector models/moondream models
face analysis / similarity detect
dataset diversity analysis - angles

[BitcrushedHeart]

• YOLO models are already supported (requires conversion to .onnx, see my yolo-2-onnx converter)
• JoyCaption Beta and Qwen3 VL are already supported.
• I'll build the other stuff into 1.1.0