10x smaller vector indexes in pgvector

pgvector is an open-source extension for PostgreSQL that adds vector similarity search, a core capability for AI applications like Retrieval-Augmented Generation (RAG), semantic search, and recommendation systems. This pull request proposes a new index type that is roughly 10x smaller than existing pgvector indexes, which is significant because current in-memory vector indexes can become prohibitively large for production deployments with millions of high-dimensional vectors. The core trade-off in vector search is between accuracy (recall), speed (latency/throughput), and memory usage — shrinking indexes by an order of magnitude makes vector search far more practical for cost-sensitive or memory-constrained environments, potentially reducing infrastructure costs and enabling use cases that were previously infeasible.