[Feature Request] PgVectorScale vs PGVector for performance

[qdrddr]

Purpose of the feature.
PgVectorScale complements pgvector, the open-source vector data extension for PostgreSQL. Please consider adding it to the documendb package.
https://medium.com/@simeon.emanuilov/fee1f9349efc

Describe the solution you'd like
Add the PgVectorScale extension to this PG package to be built-in.

Describe alternatives you've considered
Use other PG implementations with PgVectorScale built-in, add the extension manually or use slower PGVector.

Additional context
Testing it works:

1. Create a table with an embedding column:

CREATE TABLE IF NOT EXISTS document_embedding (
    id BIGINT PRIMARY KEY GENERATED BY DEFAULT AS IDENTITY,
    metadata JSONB,
    contents TEXT,
    embedding VECTOR(1536)
);

2. Populate the table with your data, including the vector embeddings. You can use the same clients and methods as pgvector for this step.
3. Create a StreamingDiskANN index on the embedding column:

CREATE INDEX document_embedding_idx ON document_embedding
USING diskann (embedding);

4. Query the table to find the closest embeddings using the index:

SELECT *
FROM document_embedding
ORDER BY embedding <=> $1
LIMIT 10;

[lichoil]

the main complements of PgVectorScale over pgvector is Binary Quantization and diskann index type support.
we will include pg_diskann developed by MS in documentDB oss, internall we've included it in our engine, but it's not opensourced, once it open sourced we will move it to oss.
Binary Quantization, we internally run benchmark, it seems its recall accuracy highly associated with data distribution. we will support PQ in diskann which can give better tradeoff of recall/rps.

so, we don't want to integration of PgVectorScale in short term.