trailblazers: Fix AI API key prompt when Writer/Rewriter polyfill falls back to native Prompt API

[tomayac]

Summary

• Same bug and fix as Fix AI API key prompt when Writer/Rewriter polyfill falls back to native Prompt API starter-extended-blog#18, applied to trailblazers/.
• The editor's AI features asked for a cloud AI API key even when the built-in-ai-task-apis-polyfills fallback was itself running entirely on-device via the browser's native Prompt API (LanguageModel, supported from Chrome 148).
• Root cause: isNative was computed as apiClass.toString().includes('[native code]'). The polyfill installs a plain JS class (class extends apiClass {}), so this check is always false for the polyfill path — even when it needs no cloud key at all.
• Replaces the toString() sniff with a shared isNativeAIFeature() helper (ai-ui-utils.js) that checks the polyfill's __isPolyfill marker and, when polyfilled, whether the underlying LanguageModel is itself native rather than also polyfilled.
• Applied consistently across all affected features: Writer, Rewriter, Summarizer, Classifier, tag suggestions, Translator, and multimodal image metadata generation.
• dist/js/ai/*.js is updated to mirror public/js/ai/*.js (eleventy passthrough-copies these files verbatim, no transform — verified byte-identical pre-fix). A full npm install/npm run build was skipped for this change since it would pull in large transitive ML runtime deps (transformers.js, onnxruntime-web, webllm) unrelated to the fix; please regenerate dist/ via the normal build/deploy flow if preferred before merging.

Test plan

• node --check on all modified files
• Manually verify in a browser without native Writer/Rewriter but with native Prompt API support (Chrome 148+) that AI actions run without prompting for an API key