llm.get_model() should return default model, make get_default_model()/set_default_model() documented APIs

[simonw]

The default model stuff is currently in llm/cli.py but it should be part of LLM core instead.

This should work:

import llm
model = llm.get_model() # Defaults to `gpt-4o-mini` or whatever was set as default
print(model.prompt("hi"))

Inspired by this change:

• OpenAI plugin should use self.get_key() #552

[simonw]

Updated documentation: https://github.com/simonw/llm/blob/6deed8f97680281433abb158fe023c2bf225d99d/docs/python-api.md