ComfyUI-INT4-Fast enables W4A4 inference on RTX 3060

Custom node brings INT4 (W4A4) model inference to ComfyUI, demonstrated on Krea2 Turbo with 6GB VRAM. Built as a standalone adaptation of existing work for ultra-fast, memory-efficient performance.