fix(responses): handle null and empty text in parse_text

[Kuldeeep18]

• I understand that this repository is auto-generated and my pull request may not be merged

Changes being requested

This PR fixes a runtime crash in structured outputs for the Responses API (client.responses.parse() and client.responses.stream(..., text_format=...)) when an output item contains null (None) or empty string ("") text.

Problem

In src/openai/lib/_parsing/_responses.py, parse_text() previously assumed text is always a valid JSON string:

if is_basemodel_type(text_format):
    return cast(TextFormatT, model_parse_json(text_format, text))

return cast(TextFormatT, TypeAdapter(text_format).validate_json(text))

If an output_text content item has text=None or text="" (e.g., from compatible third-party gateways, edge cases, or intermediate states), model_parse_json() or TypeAdapter.validate_json() raises an unhandled pydantic.ValidationError or JSON decode error instead of setting .parsed to None.

In contrast, the Chat Completions parsing helper (src/openai/lib/_parsing/_completions.py:maybe_parse_content) already safely guards against missing text:

if message.content and not message.refusal:
    message.parsed = _parse_chat_response(...)

Solution

1. In src/openai/lib/_parsing/_responses.py, updated parse_text() type signature to accept text: str | None and added an early guard:

   if not text:
       return None

2. Added comprehensive unit and streaming regression tests in tests/lib/responses/test_commentary_parsing.py covering both sync and async paths for None and "" text.

Note: Per CONTRIBUTING.md, code under src/openai/lib/ is handwritten and not modified by the code generator.

Additional context & links

• Passes all unit tests: uv run --all-extras pytest tests/lib/responses/ (434 passed)
• Passes type checks and linter: uv run ruff check, uv run ruff format --check, uv run mypy src/openai/lib/_parsing/_responses.py