Add reasoning support for custom models.

[RahulVerma989]

Currently below gives an error when using groq's deepseek r1 model. Although it works fine if I don't pass reasoning in ModelSettings. Just that I don't receive any reasoning content of the model.

return Agent(
            name="Agent",
            instructions="You are Json, an helpful agent that can delegate complex tasks to other agents.,
            handoff_description="Handoff to me.",
            model=OpenAIChatCompletionsModel(model=model, openai_client=client),
            model_settings=ModelSettings(
                reasoning=Reasoning(
                    effort="high"
                )
            ),
            hooks=self.hooks,
        )

Error Message:

{'error': {'message': "property 'reasoning_effort' is unsupported, did you mean 'reasoning_format'?", 'type': 'invalid_request_error'}}

[rm-openai]

Makes sense. Can you provide some sample code that shows how the custom model accepts reasoning parms? e.g. a curl or some sample python/