[sumimakito]

This pull request lets apps remember enrolled people in Sherpaw without their audio:

• Save an enrolled person's voiceprint with tracker.voiceprint(speaker): their embedding and the seconds of speech behind it.
• Enroll the person again in a later session with enrollVoiceprint(voiceprint).
• Add more speech to a restored person as before. It is weighted against the saved seconds.
• In the demo, known speakers stay after a reload. The browser keeps their voiceprints, not their audio. A person enrolled with another embedding model needs to enroll again.

Next: releasing @sherpaw/speaker-diarization with the other packages.

[github-actions]

✅ Deploy to Cloudflare Workers (Preview) for sandbox ready!

Name	Link
🔍 Latest deploy log	https://github.com/moeru-ai/sherpaw/actions/runs/38274682550
😎 Deploy Preview	https://eba29a00-moeru-ai-sherpaw.moeru-ai.workers.dev
🚀 Pull Request Preview	https://pr-32-moeru-ai-sherpaw.moeru-ai.workers.dev