docs(experimental): demonstrate table-first luVS benchmarks

[ibgreen-openai]

Goals

Document the table-first luVS architecture accurately and provide an opt-in, real-browser performance benchmark without introducing a second embedding storage owner.

Changes

• Document the complete Arrow FixedSizeList -> GPUVector<FixedSizeList> -> GPUTable -> borrowed graph view workflow, including explicit validity/ID siblings, padding, source batches, and ownership.
• Describe exact metrics, GPU-hashed linked selections, deterministic k-means, genuinely index-directed IVF approximation, caller-owned outputs, and RAPIDS cuVS inspiration without copied source or implied affiliation.
• Preserve upstream experimental navigation entries while adding optional luVS docs and release notes.
• Run CPU exact, GPU exact, GPU filtered, and GPU IVF-flat benchmarks over real ordinary GPUTable/GPURecordBatch columns.
• Report independent upload, index-build, command execution/fencing, readback, median timings, and recall@K.
• Cover SSR/opt-in behavior, source attribution, Float32 near-ties, and actual Chromium/WebGPU end-to-end benchmark execution.

Verification

• nvm use (Node 22.22.1).
• yarn install: the configured package registry returns HTTP 403 for required tarballs; verification reuses the existing isolated dependency tree and CI performs the normal immutable install.
• Focused documentation/SSR/attribution and real WebGPU benchmark coverage.
• yarn lint fix.
• yarn build after final source and formatting changes.
• yarn test-node.
• CI=true yarn test with real Chromium/WebGPU coverage.
• yarn website:build.
• (cd website && yarn build).

Risks

Benchmark timings depend on the browser, adapter, workload, and thermal conditions. Execution stays opt-in and gracefully reports missing WebGPU support.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 5430a29048

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".