fix(openai): handle incomplete Responses events

[dawNotPoi]

Summary

• treat response.incomplete as a terminal Responses WebSocket event
• parse the SDK event and surface its reason as a non-retryable provider error for both HTTP and WebSocket streams
• add hermetic regression coverage for transport termination, retry behavior, and response-context preservation

Fixes #7013

Testing

• uv run pytest tests/test_plugin_openai_responses.py --plugin openai -q
• uv run ruff format --check livekit-plugins/livekit-plugins-openai/livekit/plugins/openai/responses/llm.py tests/test_plugin_openai_responses.py
• uv run ruff check livekit-plugins/livekit-plugins-openai/livekit/plugins/openai/responses/llm.py tests/test_plugin_openai_responses.py
• uv run mypy --platform linux -p livekit.plugins.openai
• git diff --check

[CLAassistant]

All committers have signed the CLA.

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no bugs or issues to report.

[gabriele-dialpad]

Thanks for fixing this!

[chenghao-mou]

thanks for the PR!