Right now, the serve context can be used to run the inference server with explicit control via the server_args kwarg. It would be helpful to include explicit arguments for this behavior in the function itself to simplify the API for this behavior since there's lots of use cases where it would be helpful (especially in light of #3)