CodSpeed micro-benchmarks for the binding hot paths (produce, ingest, UDF). Design rationale: PLAN.md. CI: ../.github/workflows/codspeed.yml.

Every benchmark carries exactly one (registered in conftest.py):

• gate: binding-dominated, GIL-held, deterministic under Callgrind. A threshold breach is a binding regression.
• informational: engine/library/streaming-diluted. Reported, never gated (would false-positive on engine bumps).

Only walltime runs locally (no Valgrind on macOS arm64; instruction-count gating is Linux/CI-only, and walltime is noisy on sub-ms benches). Pin the data libs identically across both builds so the delta is pure binding:

for P in ../main/.venv-release/bin/python .venv-release/bin/python; do
  $P -m pytest benchmarks/<module>.py --codspeed --codspeed-mode=walltime -o addopts= -p no:cacheprovider
done

• READ aggregates real columns (sum/length), never count(*) (answered from metadata).
• WRITE fully materializes the result or drains the lazy reader.
• Warm once before measuring.
• con fixture pins threads=1 (see conftest.py).

Two traps (a benchmark that skips these silently measures the wrong thing):

• OUT-col null benches need REAL nulls (CASE WHEN ... THEN NULL), else the cheap std::move path is taken.
• IN-numpy string benches need mixed ASCII + non-ASCII + a null sentinel, else the transcode/null ladder is skipped.