Fixes custom model parameter overrides in config

[JasonWeill]

Removes model_parameters, which were redundant with provider_params, from chat and completion code. Fixes #624.

Tested by passing a custom model parameter in my JupyterLab config:

jupyter lab --AiExtension.model_parameters openai-chat:gpt-4-1106-preview='{"max_tokens": 4095}'

In the chat UI and the completion interface, this model now works. Temporarily added log statements confirm that the custom parameter is included in the llm object exactly once.

model_parameters was not redundant with provider_params in the magic command handler, so it is not modified. Magic commands with a model with custom model config continue to work.

[dlqqq]

This breaks traitlets configurability because this PR just deletes model_parameters, the trait being passed from the top-level AiExtension class in extension.py:

model_parameters = Dict(
    key_trait=Unicode(),
    value_trait=Dict(),
    default_value={},
    help="""Key-value pairs for model id and corresponding parameters that
    are passed to the provider class. The values are unpacked and passed to
    the provider class as-is.""",
    allow_none=True,
    config=True,
)

Please find another way to fix this bug that doesn't delete the traitlets configuration.

[JasonWeill]

Closes in favor of #632, which was approved.