Reproduces app.py's TeleStyleV2 style-transfer pipeline as native ComfyUI
nodes (Qwen TeleStyle Loader + Qwen TeleStyle Apply), loading weights
directly from ComfyUI's own model folders instead of bridging live
MODEL/CLIP/VAE objects (which doubled memory and OOM'd on this box).
