Improve support of Reference-only and outpainting for SDXL

Reference-only has shown be a very powerful mechanism for outpainting as well as image variation. We should investigate a bit how we can best support this in a modularized, library-friendly way in `diffusers`. Also I think we should try this out for SDXL.

Related links:
- https://github.com/Mikubill/sd-webui-controlnet/discussions/1280
- https://github.com/Mikubill/sd-webui-controlnet/discussions/1464
- https://github.com/Mikubill/sd-webui-controlnet/discussions/1236

and

https://github.com/huggingface/diffusers/blob/main/examples/community/stable_diffusion_controlnet_reference.py
https://github.com/huggingface/diffusers/blob/main/examples/community/stable_diffusion_reference.py

@DN6 maybe this could be a first fun project for you