Fixes tool_choice handling in the Python SDK `Responses API

[maryamfaizan53]

Summary

Fixes tool_choice handling in the Python SDK Responses API.
Previously, only tool_choice="auto" was accepted. Structured tool choices (e.g. {"type": "code_interpreter"}) were ignored.

Changes

• Updated ResponsesCreateParams.tool_choice to accept both Literal["auto"] and Dict[str, Any].
• Added test test_tool_choice_object in tests/test_responses.py.

Why

• Documentation states that structured tool_choice is supported:
  • https://platform.openai.com/docs/guides/latest-model#allowed-tools
  • https://platform.openai.com/docs/api-reference/responses/create#responses_create-tool_choice
• Without this fix, tool_choice always defaults to "auto".

Result

Now both of these work:

tool_choice="auto"
tool_choice={"type": "code_interpreter"}