Inspect integration removes model reasoning

[wichersn]

The inspect integration calls renderers.get_text_content on the model response here, which removes the model reasoning before passing onto inspect.

This is an issue because users may want to create inspect evals which depend on the model's reasoning. It's also helpful to be able to see the model reasoning in the inspect UI for debugging.

Expected behavior:
Both the reasoning and response parts should be passed to the InspectAIChatMessageAssistant as ContentReasoning and ContentText. Then evals can use the reasoning, and the reasoning will be displayed in the inspect UI.