add vector search benchmarks

[vkarpov15]

What this PR does:

Basic benchmarks for vector search:

1. writeVector: insert 21k vector documents in batches of 20 using insertMany with ordered: false
2. queryVector: execute 2000 ANN queries 10 at a time on the 21k vectors inserted

21k documents just have a text and vector property. Paradoxically, it seems like collections are actually faster than tables in this case. A big reason why collections are faster though is because collections project out the vector field by default, it seems like nearly 90% of the overhead of tables vs collections is eliminated by projecting out the vector field on the read path.

Tables

• Tables insert: {"inserted":21347,"seconds":376.129264,"docsPerSecond":56.754425,"secondsPerBatch":0.352395}
• Tables read: {"queries":2000,"parallelism":10,"seconds":55.601865,"queriesPerSecond":35.970016,"secondsPerBatch":0.278009}
• Tables read w/ vector projected out: {"queries":2000,"parallelism":10,"seconds":32.994885,"queriesPerSecond":60.615457,"secondsPerBatch":0.164974}

Collections

• Collections insert: {"inserted":21347,"seconds":222.619178,"docsPerSecond":95.890211,"secondsPerBatch":0.208572}
• Collections read: {"queries":2000,"parallelism":10,"seconds":28.87107,"queriesPerSecond":69.273497,"secondsPerBatch":0.144355}

Which issue(s) this PR fixes:
Fixes #

Checklist

• Changes manually tested
• Automated Tests added/updated
• Documentation added/updated
• CLA Signed: DataStax CLA

[Copilot]

Pull request overview

Adds basic vector search benchmark scripts to compare insert and query performance between Astra tables and collections.

Changes:

• Introduces benchmarks/writeVector.mjs to create/sync a content table or collection, then bulk-insert ~21k vector documents in batches and log throughput.
• Introduces benchmarks/queryVector.mjs to run 2,000 parallelized ANN-style vector queries against the content model and log query performance metrics.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 5 comments.

File	Description
benchmarks/writeVector.mjs	Sets up the Content schema and collection/table, then bulk-inserts vectorized movie documents and prints insert performance stats.
benchmarks/queryVector.mjs	Uses the existing Content schema to perform repeated vector similarity queries and prints query performance stats.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[vkarpov15]

I checked and confirmed that writing $binary doesn't meaningfully help the writeVector case