Add MiniMax text-to-image support

[octo-patch]

Reason: Add direct MiniMax text-to-image generation to the executable image-creator skill.

• add global and China endpoint selection with Bearer authentication
• support both image models and the documented text-to-image request controls
• decode URL and base64 responses with API error handling
• document setup and usage, with focused provider tests

Checks:

• PYTHONDONTWRITEBYTECODE=1 uv run python -m unittest discover -s .claude/skills/ai-image-creator/tests -v
• PYTHONDONTWRITEBYTECODE=1 uv run python .claude/skills/ai-image-creator/scripts/generate-image.py --list-models
• PYTHONDONTWRITEBYTECODE=1 uv run python -c 'from pathlib import Path; p=Path(".claude/skills/ai-image-creator/scripts/generate-image.py"); compile(p.read_text(), str(p), "exec")'
• git diff --check

[octo-patch]

I added support for MiniMax subject references via repeatable --subject-reference HTTPS URLs and updated the focused tests and docs. I ran python3 -m unittest discover -s .claude/skills/ai-image-creator/tests -v, the model-list smoke check, the compile check, and git diff --check.