feat(image-gen): add OpenAI Codex image edit provider

[BeliefanX]

Summary

• Adds prompt-guided image editing support to the OpenAI Codex/ChatGPT-auth image provider.
• Sends ordered reference images through the Codex Responses image_generation tool with action: "edit".
• Validates data URLs and constrains local file references to Hermes image-cache roots before upload/forwarding.

Relationship

• Layer 3 of the image-edit split: concrete OpenAI Codex provider support.
• Logical review order: feat(image-gen): add provider edit contract #26967 → feat(image-gen): validate image references #26968 → feat(image-gen): add OpenAI Codex image edit provider #40491 → feat(image-gen): add opt-in image_edit tool #40498.
• These PRs are intentionally related but not GitHub-stacked; each targets main so CI remains independent.
• Supersedes the provider portion of feat(image-gen): add OpenAI image edit provider #26969, which is closed in favor of this cleaner provider-only replacement.
• Tool registration is intentionally kept out of this PR and split into feat(image-gen): add opt-in image_edit tool #40498.

Testing

• venv/bin/python -m py_compile plugins/image_gen/openai-codex/__init__.py tests/plugins/image_gen/test_openai_codex_provider.py
• venv/bin/python -m pytest tests/plugins/image_gen/test_openai_codex_provider.py -q -o 'addopts=' — 28 passed
• custom static guard confirmed no sort/set/dedupe in the reference collection path
• git diff --check

[BeliefanX]

Superseded by the merged unified image-to-image/Codex implementation in #48705 and its follow-up Codex input support; closing this stale conflicting branch.