Parallelize verifier pairing inputs under std

[Neotamandua]

Summary

This change parallelizes the construction of the two independent G1 pairing inputs during proof verification using rayon via joining two closures.

The left opening-witness computation and the right grouped MSM are independent, so they can execute concurrently under std. The existing non-parallel behavior is retained for no_std.

This does not change the verification equations, transcript, proof format, or public API.

A Criterion verification benchmark using a synthetic circuit with 32 public inputs is added as well.

Performance

The timings below were measured separately using an existing valid Phoenix V3 fixture with 32 public inputs (not the synthetic benchmark added in this PR), with verifier/proof decoding outside the timed section.

	Median
Baseline	2.915 ms
Parallel	2.057 ms
Improvement	29.45%

3 paired runs were performed (25 warmups and 500 timed verifications each), using the default 32-worker Rayon pool.

This measures single-proof verification latency.

[codecov]

Codecov Report

❌ Patch coverage is 71.42857% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 92.25%. Comparing base (b932820) to head (9ad817f).
⚠️ Report is 3 commits behind head on master.

Files with missing lines	Patch %	Lines
src/proof_system/proof.rs	71.42%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1000      +/-   ##
==========================================
- Coverage   92.27%   92.25%   -0.03%     
==========================================
  Files          81       81              
  Lines        8886     8887       +1     
==========================================
- Hits         8200     8199       -1     
- Misses        686      688       +2     

Files with missing lines	Coverage Δ
src/proof_system/proof.rs	96.94% <71.42%> (-0.36%)	⬇️

---

Continue to review full report in Codecov by Harness.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b932820...9ad817f. Read the comment docs.

🚀 New features to boost your workflow:

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

[moCello]

COMMENT — The verifier change is correct: both closures read only public, already computed values, the pairing equation is unchanged, and the split matches the rayon::join pattern in the prover. Approval waits on the new bench, which cannot run from this repo and makes make bench fail.

One small note below on the V1 path.

[moCello]

APPROVE — Both verifier paths compute the two pairing inputs with rayon::join, as the prover already does, and the pairing equation does not change. The closures read only public values that the transcript fixes, the no_std path stays serial, and the new bench runs from this repository.

• Process note: the pull request description still says that a Phoenix V3 benchmark is added, and its timing table comes from that external fixture. The bench in this change is a synthetic circuit with 32 public inputs. Update the description so that it names the bench that lands.