Add optional video input to the chat completions input path

[octo-patch]

Reason: the chat-completions input path builds only text and image content parts, so models that also accept video input cannot receive it.

Summary

• Add an optional, keyword-only video_url argument to generate_text, generate_stream_text, and structured_response, mirroring the existing image_url argument.
• Append a {"type": "video_url", "video_url": {"url": ...}} content part only when the caller passes a video, so text-only and image requests keep their current payload.
• Add offline tests that assert the content parts sent for text-only, image, video, and combined image plus video inputs across all three methods.

Notes

• The change is additive: no existing signature, default or request payload changes when video_url is not passed.
• Providers that subclass this compatible chat path inherit the video input capability without further changes.
• The tests stub the client through the cached properties, so they need no API key and make no network calls.

Checks

• ruff check simplemind/providers/openai.py tests/test_input_modalities.py (all checks passed)
• ruff format --diff tests/test_input_modalities.py (1 file already formatted)
• python3 -m compileall -q simplemind tests (passed)
• python3 -m pytest -q tests/test_input_modalities.py tests/test_basics.py (8 passed)
• python3 -m pytest -q --collect-only tests (36 tests collected)