Hi, I am looking at this: https://github.com/bentoml/Yatai/issues/156, feat: support async endpoint Is there a way to create something similar with BentoML + something? For example, when serving Stable Diffusion models, it may take some seconds to do this.