Retrain PuLID for Z-Image

[artyfacialintelagent]

Z-Image is a new model that focuses on realism and is half the size of Flux dev (only 6B parameters), but with image quality and prompt adherence better than Flux and almost as good as Qwen. The small size means that Z-Image is easy to train - a LoRA can be trained in just a couple of hours on a consumer GPU.

https://github.com/Tongyi-MAI/Z-Image

That should mean that it is easy to retrain PuLID for Z-Image too. @ToTheBeginning, please consider adding Z-Image support!