docs: update build performance benchmarks

[github-actions]

Automated build benchmark update from workflow run #57.

[greptile-apps]

Greptile Summary

Automated benchmark update for v2.4.0 that adds new performance data to BUILD-BENCHMARKS.md and updates the Performance section in README.md.

Key changes:

• Native build speed regressed from 1.9 to 2.2 ms/file (+16%)
• WASM build improved slightly from 6.6 to 6.4 ms/file (-3%)
• File count increased from 99 to 109 files
• DB size per file increased from 3848 to 4397 bytes (+14%)

Critical issue:

• The "Lightweight Footprint" subsection (h3) was unintentionally deleted from README.md. This is caused by a regex bug in scripts/update-benchmark-report.js:220 that captures h3 subsections when replacing the Performance section. The automation script needs to be fixed to preserve subsections.

Confidence Score: 2/5

• Not safe to merge - unintended documentation removal
• While the benchmark data was updated correctly, the PR unintentionally removed the "Lightweight Footprint" section from README.md due to a bug in the automation script, resulting in lost documentation content
• README.md requires attention - the removed "Lightweight Footprint" section must be restored

Important Files Changed

Filename	Overview
README.md	Updated benchmark numbers (native build regressed 16%), but unintentionally removed "Lightweight Footprint" subsection
generated/BUILD-BENCHMARKS.md	Added v2.4.0 benchmark data correctly with trend indicators

_{Last reviewed commit: cef12b1}

[greptile-apps]

_{2 files reviewed, 1 comment}

_{Edit Code Review Agent Settings | Greptile}

[greptile-apps]

The "Lightweight Footprint" subsection (### heading with dependency table) was unintentionally removed. This is caused by a bug in scripts/update-benchmark-report.js line 220 — the regex /## 📊 Performance\r?\n[\s\S]*?(?=\r?\n## (?!#)|$)/ captures h3 subsections when replacing the Performance section.

The script should only replace the benchmark table, not the entire section including subsections.

[carlos-alm]

Superseded by #121 which includes expanded benchmark format (incremental + query metrics).