-- pg_turbovec v1.9.0 -- oversampling + testing/bench hardening.
--
-- This file is a *reference mirror*. The authoritative install
-- script is generated by `cargo pgrx schema`.
--
-- v1.9.0 adds one user-facing feature plus testing/benchmark
-- infrastructure on top of v1.8.0:
--
--   1. turbovec.oversample GUC (float, default 1.0, range
--      1.0..=100.0): the scan fetches ceil(search_k * oversample)
--      quantized candidates and the executor's reorder queue trims
--      to the exact top-k. Widening the candidate set recovers true
--      neighbours the lossy quantized ranking placed just outside
--      search_k -- turning quantization into a tunable recall
--      frontier (matching Qdrant oversampling / VectorChord rerank).
--      Default 1.0 = identical to v1.8.0 behaviour.
--
--   2. Test coverage: medium-scale (20k-row) recall-floor tests per
--      bit_width + distinct-id assertions on every ANN-scan test
--      (the guard that would have caught the pre-AVX2 bug). See
--      docs/TESTING.md.
--
--   3. First published head-to-head benchmark (docs/BENCHMARKS.md +
--      benches/scripts/vectordbbench/): Cohere wikipedia 1M x 1024-d
--      vs pgvector HNSW. Confirms recall@10 = 1.000 on the fixed
--      v1.8.0+ build; storage 7.6x (4-bit) / 15.2x (2-bit) smaller.
--
-- The on-disk relfile format is byte-identical to v1.7.x / v1.8.x
-- (MetaPageData::version = 3). The new GUC is additive and defaults
-- to no-op (1.0). There is no on-disk migration. The upgrade path
-- `ALTER EXTENSION pg_turbovec UPDATE TO '1.9.0';` resolves cleanly.
-- No REINDEX needed.