Additional minimal changes for VDB modular backend support @https://github.com/mlcommons/storage/pull/316

[idevasena]

Summary

This PR documents the new modular backend-agnostic VectorDB benchmark runner as a standalone preview and adds the supporting fixes/tests needed to validate it independently from the existing mlpstorage vectordb path.

The modular runner is available through:

python -m vdbbench.benchmark

or

uv run python -m vdbbench.benchmark

Scope of this PR

This PR includes the following minimal changes:

Documentation

1. Updates vdb_benchmark/README.md to describe the modular runner as a standalone preview.
2. Documents how to invoke the modular runner directly.
3. Documents backend dependency requirements.
4. Documents backend connection environment variables.
5. Clearly states that multi-node and multi-client VDB benchmark execution is not implemented yet.
6. Clarifies that ./mlpstorage vectordb continues to use the Milvus-oriented runner.
7. Adds tests for different backend support

[github-actions]

MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅

[russfellows]

PR #379 — Additional minimal changes for VDB modular backend support
Author: idevasena | +8787 / -428 | 41 files | Status: dirty (merge conflicts)

Assessment: ⚠️ NOT READY — merge conflicts, and scope warrants careful review. - mergeable_state: "dirty" — the branch needs a rebase against current main before it can merge. The branch base is 258483b (pre-PR #378); main is now at c43da24, which is ~6 commits ahead. - At +8787 / -428 across 41 files this is a large PR. The stated scope is "documentation + tests for standalone python -m vdbbench.benchmark modular runner", but the diff size suggests more substantial changes. - This PR builds on PR #316 which was a large VDB modular backend refactor. Should confirm #316 is actually merged before this one. - Recommendation: Author needs to rebase on current main first; then a more detailed file-by-file review is warranted given the size.