FuseAnyPart: Diffusion-Driven Facial Parts Swapping via Multiple Reference Images