feat(api): refine image request validation

[typelicious]

What changed

• add shared request normalization for image generation, image editing, and image dry-run previews
• validate prompt, model, n, size, and scalar image fields before provider calls
• make image preview and runtime endpoints fail fast with clear invalid-request responses
• add regression tests and README/changelog updates

Why

Image routing already existed, but request validation was split across paths and allowed malformed image fields to fall through too late. This makes the image workflow stricter and more predictable.

How verified

• PYTHONPYCACHEPREFIX="$PWD/.pycache" python3 -m compileall foundrygate tests
• PYTHONPATH=. ./.venv-check-313/bin/pytest -q
• ./.venv-check-313/bin/ruff check .
• ./.venv-check-313/bin/ruff format --check .
• /usr/bin/git diff --check