valiant 0.1.0.1

Changed

• Switched the per-connection prepared-statement cache used by Valiant.Execute and Valiant.Batch from a HashPSQ-based LRU to SIEVE (NSDI'24). In standalone microbenchmarks, SIEVE is 2.3-3.8x faster than the previous LRU across fill, hit-only, mixed, eviction-storm, uniform-random, and Zipf-skewed workloads. No public API change.
• Updated pg-wire lower bound to ==0.2.* (required to pick up the SIEVE-backed Connection and the PgWire.Cache.Sieve exports).

Removed

• psqueues dependency.

Hackage