Embeddings normalization fixes

[hawkeye217]

Proposed change

For the Jina AI models, text-text cosine similarity is normally larger than text-image cosine similarity. This PR applies normalization for all vector based searches, uses cosine similarity on image-image searches, and only saves Z score normalization stats for multi-modal searches.

• Use cosine distance instead of euclidean distance for vector tables (requires users to reindex embeddings)
• Ensure we fetch model state on Explore page load

Type of change

• Dependency upgrade
• Bugfix (non-breaking change which fixes an issue)
• New feature
• Breaking change (fix/feature causing existing functionality to break)
• Code quality improvements to existing code

Additional information

Users running the dev builds will have to set reindex: True in their config to regenerate their embeddings as distance function has now been changed to cosine.

Checklist

• The code change is tested and works locally.
• Local tests pass. Your PR cannot be merged unless tests pass
• There is no commented out code in this PR.
• The code has been formatted using Ruff (ruff format frigate)

[netlify]

✅ Deploy Preview for frigate-docs canceled.

Name	Link
🔨 Latest commit	1165c42
🔍 Latest deploy log	https://app.netlify.com/sites/frigate-docs/deploys/6709679ce5cdbf0008ae64ab