Automatic CI testing of the `llm` CLI

[philpax]

While fixing #350, I've noticed that there are a few regressions in the LLM CLI. We should figure out a way to automatically run integration tests on the CLI to help reduce this.

[LLukas22]

We could pull the model downloader from the test package into the cli package and enable loading models from an url. The we just need a test harness for the cli which we can use in our CI.

[philpax]

I figured an even easier solution would be to call the CLI from llm-test, because the model should already be available. I'm not sure about downloading the model in the CLI - that seems like it might be too much magic, especially because the user won't know where it ends up.