docs(experimental): demonstrate table-first luVS benchmarks

[ibgreen-openai]

Goals

Add an opt-in interactive browser benchmark for table-backed exact and approximate GPU similarity search.

Changes

• Embed a real WebGPU benchmark in the existing conceptual luVS reference.
• Compare independent CPU exact, GPU exact, filtered GPU exact, and IVF-flat search.
• Report upload/index-build timing, fenced query latency, candidate counts, reusable index bytes, and recall@K.
• Preserve server-rendering safety, explicit device creation, Float32-equivalent tie handling, and real-browser coverage.
• Add release notes for fixed-size GPU table columns, Arrow ingestion, and luVS.

Previously reviewed as #2950. Merge after clustering and IVF reach master.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 38dfc85c10

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".