Support custom models

[WoosukKwon]

We need to provide clean abstractions and interfaces so that users can easily plug in their custom models.

[hmellor]

Can this be considered done now that there is a documentation page detailing how to do it https://docs.vllm.ai/en/latest/models/adding_model.html?

[DarkLight1337]

We now support automatic detection of generation/embedding task (vllm/models/interfaces_base.py) as well as additional features like multimodality and PP (vllm/models/interfaces.py) for both built-in and OOT models. It's safe to say that this issue has been resolved.