perf: serial threshold in lifted build_next_layer

[odinfree]

What

crates/stwo/src/prover/backend/simd/blake2s_lifted.rs, build_next_layer: use the serial chunk iterator below 2^10 nodes instead of dispatching rayon for the small fold layers. The pristine code dispatches par_chunks_mut down to single-node layers; Plonky3 uses the same 1024-node serial threshold.

The per-chunk work is refactored into a work closure shared by the parallel and serial arms; behavior is identical either way.

Why

• Neutral on large-cache hosts (measured: 32.3 vs 32.4 ms medians on a 96-column lifted trace commit at log 20, Apple Silicon) — included for hosts where dispatch dominates the tiny folds.
• Zero risk: output layers are bit-identical; committed roots unchanged.

Verification

• cargo +nightly-2026-01-15 test -p stwo --release --features prover,parallel --lib: 267/267 pass (and with --features prover).
• cargo +nightly-2026-01-15 clippy -p stwo --features prover,parallel -- -D warnings: clean (note: &closure trips needless_borrows_for_generic_args on this nightly — the closure is passed by value).
• cargo fmt --check: clean.

Stacked after nothing — independent of #1458.

[cursor]

PR Summary

Low Risk
Performance-only scheduling change with identical per-chunk hashing; existing CPU/SIMD parity tests cover Merkle roots.

Overview
build_next_layer in the SIMD lifted Blake2s Merkle path no longer always uses Rayon when the parallel feature is on. Layers with log_size below 10 (fewer than 1024 parent nodes) now fold with a serial chunks_mut loop; at log_size >= 10 it still uses par_chunks_mut, matching Plonky3’s 1024-node serial threshold.

The per-chunk SIMD Blake2 compress/transpose logic is unchanged but moved into a shared work closure invoked from either the parallel or serial iterator, so hashing output stays the same while avoiding Rayon overhead on tiny upper Merkle layers.

^{Reviewed by Cursor Bugbot for commit 6dd7dd2. Bugbot is set up for automated code reviews on this repo. Configure here.}

[reviewable-StarkWare]

This change is 

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.77%. Comparing base (826591c) to head (6dd7dd2).

Additional details and impacted files

@@           Coverage Diff           @@
##              dev    #1459   +/-   ##
=======================================
  Coverage   91.76%   91.77%           
=======================================
  Files         127      127           
  Lines       15173    15177    +4     
  Branches    15173    15177    +4     
=======================================
+ Hits        13924    13928    +4     
  Misses       1180     1180           
  Partials       69       69           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[odinfree]

Withdrawing for the time being — the author needs to run an internal review before this is considered for submission. Will resubmit (or a maintainer can pick the idea up) after that review. Thanks for the CI run.