Performance: parallelising kimchi prover more

[volhovm]

A bunch of performance improvements to the kimchi prover aiming to utilise more parallelism where available.

As of now, the save-up seems to be as follows:

TODOs:

• Check how much faster it is on mina side. Adjust the benchmark if necessary (more columns? lookups?)
  • There is a 7% improvement as opposed to the 22% on just prover side. Maybe the benchmark circuit is too simplistic (not enough lookups?). Maybe it's because mina runs some other threads in parallel (like two provers in parallel?).
• Make sure there are no regressions for the verifier side.
  • Parallelising constraints seems to add 3% to the verifier for 1024 gates (and nothing when there are 16k gates?)?? generally this is hard to verify: even running every verifier benchmark for 5 minutes I still get sometimes 0.5% noise. Really hard to tell.
• Check if parallel evaluations can be faster -- do we need to clone the vector and then interpolate?
• Check that perm_aggreg optimisation preserves correctness
• Last step: run the threading profiler once again to make sure I didn't miss anything

[codecov]

Codecov Report

Attention: Patch coverage is 97.11934% with 7 lines in your changes missing coverage. Please review.

Project coverage is 76.87%. Comparing base (22a9862) to head (5472fed).
Report is 77 commits behind head on master.

Files with missing lines	Patch %	Lines
kimchi/src/prover.rs	84.78%	7 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2969      +/-   ##
==========================================
- Coverage   76.95%   76.87%   -0.08%     
==========================================
  Files         261      261              
  Lines       61457    62092     +635     
==========================================
+ Hits        47293    47736     +443     
- Misses      14164    14356     +192     

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