feat(experimental): add table-backed GPU k-means and IVF-flat

[ibgreen-openai]

Goals

Add deterministic WebGPU k-means clustering and truly index-directed IVF-flat similarity search.

Changes

• Implement deterministic bounded Lloyd iterations without Float32 atomic additions.
• Build caller-owned IVF centroids, list counts/offsets, stable source IDs, and separate logical row references.
• Probe indexed lists directly, preserve source-aligned masks, support approximation/fallback controls, and exactly rerank selected vectors.
• Include the complete 500-line conceptual luVS reference covering architecture, motivation, ownership, worked IVF arrays, accuracy/memory tradeoffs, graph reuse, and device-loss recovery.
• Add focused Node and real-WebGPU clustering/index coverage.

Previously reviewed as #2949. Merge after exact similarity search reaches master.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 1827468355

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".