feat(wiki): add OpenAI-compatible visual fact extraction

[MarinaMackay]

Summary

Fork-side stacked PR for the PR3 slice of sysevol-ai#655. This adds an optional OpenAI-compatible VLM extractor for the visual fact contract introduced in PR2.

This is intentionally stacked on codex/mmwiki-pr2-core-knowledge-v2 in my fork so the diff only shows the VLM extractor slice until upstream PR2 lands.

Changes

• Add OpenAICompatibleVisualFactExtractor for OpenAI-compatible chat/VLM endpoints.
• Attach repository media as bounded data URLs when a local repo path is provided.
• Normalize model output through the existing visual fact contract.
• Validate API base URLs, API keys, model/provider labels, timeouts, MIME types, repository-relative artifact paths, response bytes, and image bytes.
• Add focused tests for provider requests, normalization, unsafe configuration, unsafe artifact paths, MIME rejection, response bounds, image bounds, and invalid JSON.

Type of Change

• Bug fix
• New feature
• Breaking change
• Documentation update
• Refactoring
• Performance improvement
• Tests

Testing

• Tests pass locally
• Added new tests for the changes

Commands run locally:

python -m pytest test/wiki
python -m pytest test/wiki/test_media_vlm.py test/wiki/test_media_facts.py
python -m py_compile codenib/wiki/media_vlm.py codenib/wiki/media_facts.py codenib/wiki/__init__.py
git diff --check codex/mmwiki-pr2-core-knowledge-v2...HEAD

Local result:

387 passed

Checklist

• My code follows the project's style guidelines
• I have performed a self-review of my code
• I have commented my code, particularly in hard-to-understand areas
• My changes generate no new warnings
• Any dependent changes have been merged and published

Notes

Do not open this slice upstream until sysevol-ai#665 lands or maintainers ask for parallel upstream drafts. Opening it upstream now would include cumulative PR2 changes.

[fishmingyu]

PR2 has merged upstream as sysevol-ai#665. I restacked this PR3 slice directly onto the resulting upstream main, preserved the original feature commits and authorship, added a maintainer provenance/media-integrity hardening commit, and opened the focused upstream PR as sysevol-ai#668. This fork-side comparison draft can now be closed when convenient.