Get gpt4all to use GPU

[BWShor]

I'm new to github and don't know where to post this but I was able to get gpt4all to use GPU as well as the "Reasoner v1" model by making the following changes to jupyter-ai-magics/providers.py:
add the setdefualt("device"... argument after the n_threads argument is added:
kwargs["n_threads"] = max(int(n_threads), 1)
kwargs.setdefault("device", "kompute")

Add the qwen2.5-coder-zb-instruct-q4_0 line at the end of the models list:
"qwen2.5-coder-7b-instruct-q4_0",
# "qwen2-1_5b-instruct-q4_0" - I'm still tinkering with the 1_5b-instruct model and haven't actually tested it

If this is useful to anyone, please let me know!

[srdas]

@BWShor Contributions that enhance Jupyter AI's use of models like gpt4all are welcome -- after you are done testing, feel free to open a pull request with the modified code for review. Provide as much detail as you can in the PR, such as screenshots, videos, explanation of the change with testing you did, etc.