docs(blog): add Curvine and LanceDB benchmark

[LJF-050]

Summary

Add a bilingual blog post covering the LanceDB benchmark on Curvine.

The benchmark compares three storage paths:

• OSS for both data and indexes
• OSS data with indexes cached by Curvine
• Curvine for both data and indexes

Key results include:

• Vector-search p50 latency reduced from 11 ms to 2 ms
• Vector-search QPS increased from 69.4 to 333.3
• Vector-search p99 latency reduced from 62 ms to 10 ms
• Full-text and hybrid search also showed significant latency and throughput improvements
• Filtered vector search showed limited gains because its bottleneck was scanning and computation rather than index I/O

Changes

• Add the English blog post
• Add the corresponding Chinese translation
• Add the LanceDB × Curvine benchmark image to both locales
• Format benchmark results as Markdown tables
• Document index-only and full-cache adoption strategies

Validation

• Ran npm run build
• English production build passed
• Chinese production build passed
• Verified both preview pages return HTTP 200
• Confirmed the article image renders in both locales

Files

• blog/2026-07-31-curvine-lancedb-vector-search/
• i18n/zh-cn/docusaurus-plugin-content-blog/2026-07-31-curvine-lancedb-vector-search/