Automatically launch a server with a default model

[davidrpugh]

I think it should be possible to use jupyter-server-proxy to configure a llama-server so that it can automatically launched when a user launches JupyterLab.