[xwj-ox]

• Does this PR have a descriptive title that could go in our release notes?
• Does this PR add any new dependencies?
• Does this PR modify any existing APIs?
• Is the change to the API backwards compatible?
• Should this result in any changes to our documentation, either updating existing docs or adding new ones?

Reference Issues/PRs

Stacked on #1536 (u/xiajiang/ivf-reassignment). This is the consumer at the top of the IVF stack.

Review order: #1534 -> #1535 -> #1520 -> #1536 -> #1519.

What does this implement/fix? Briefly explain your changes.

Add ivf-index to the shared benchmark runner with typed build/search inputs, a private in-memory provider, phase metrics, checkpoints and shared recall/QPS reporting. Preserve FP16/FP32 bin and fvecs loading, supplied groundtruth, build-only runs and the independent full-probe oracle. The baseline and projection-gemm JSON presets compose the independent library policies; default recall remains Recall@1000.

Any other comments?

The diff is confined to diskann-benchmark. Validation: six benchmark/integration tests passed with default features and six with both IVF features, including the supplied-truth regression. Four release CLI runs (baseline/combined, one/four threads) used 4,096 synthetic vectors and all 128 queries at Recall@1000; full-probe checks passed. Workspace and benchmark feature-matrix Clippy with warnings denied, formatting and license-header checks passed. This reports in-memory IVF-Flat behavior and does not establish a performance improvement.

The prerequisite already fails invalid_initialization_and_uninitialized_insertion_fail because NaN initial centroids are accepted. This was reproduced on unmodified base 83c0109 and remains the only failure in the full IVF suite.