Fix Python SDK config compatibility

[ryanbaumann]

Summary

Fix the portable infographic skill on Python 3.9, where the latest compatible google-genai release (1.47.0) does not expose the newer typed image_size or thinking_level fields.

Changes

• Feature-detect SDK configuration fields before constructing typed image and thinking configs.
• Preserve explicit resolution and HIGH thinking on newer SDKs.
• Fall back to native image resolution and dynamic thinking on google-genai 1.47.0 instead of failing validation before the API request.
• Align the supported boundary and installer on Python 3.9+ and google-genai>=1.47.0.
• Add regression tests for the 1.47 and current typed-config surfaces.
• Run the Python regression in normal CI and the npm publish sanity check.
• Prepare patch release infographic-agent@3.2.1.

Root cause

The Gemini API supports the requested image size and high thinking settings, but older Python SDK type models reject those fields during local validation. Python 3.9 currently resolves to google-genai 1.47.0, so generation and refinement failed before reaching Gemini.

Verification

• npm run lint — passed
• npm test — 68 passed, 9 skipped
• python3 -m unittest skill/infographic-agent/test_portable_infographic.py — 2 passed
• npm run build — passed
• Publish sanity: Node syntax, Python compile/tests, and npm pack --dry-run — passed
• Independent review — clean after two compatibility-boundary corrections

Release note

Merging this PR triggers the repository's publish workflow. Because 3.2.1 is not currently on npm, the workflow will publish and tag infographic-agent@3.2.1 automatically.