[prrao87]

Why

The benchmark tables in the README and docs showed lance-csr 0.2.0, one release behind. The benchmark repositories now pin 0.3.1 and have new lance-csr runs, merged in graph-benchmark-ldbc#32 and graph-benchmark-ldbc-complex#3. This PR syncs the tables from those runs.

What changed

• The tables, summaries and notes now come from the 0.3.1 runs. Over the 14 IC queries, lance-csr in memory goes from 2.8× to 3.9× faster than Neo4j, mostly from IC12 and IC13. The 30-query summary moves by less than 10%, though Q30 itself is about 20% slower than in 0.2.0.
• Only lance-csr was re-run, so the notes now say when lance-csr and the other engines were measured, instead of claiming every engine ran one after another. That claim still appears when a spec has no others_measured date.
• The notes say when lance-csr ran on battery power, which can throttle the CPU, if the benchmark repository records it. These runs were on mains power, so the current notes don't mention it.
• The spell check now ignores git commit hashes, since 6a582ba failed it.

Validation

• Ran docs/sync_benchmarks.py against the pinned commits and read the regenerated notes.
• uvx typos@1.26.0 ., uv run --project docs --locked zensical build --strict and ruff check/ruff format --check on the sync script passed locally.
• The tables are pinned to the merge commits of both benchmark PRs, now on their main branches.

🤖 Generated with Claude Code

[cloudflare-workers-and-pages]

Deploying lance-csr with    Cloudflare Pages

Latest commit:	0f09807
Status:	✅  Deploy successful!
Preview URL:	https://99db3bee.lance-csr.pages.dev
Branch Preview URL:	https://docs-benchmarks-lance-csr-0.lance-csr.pages.dev

View logs