Better parallelization

[ed255]

The following image is the CPU utilization of a circuit with ~1000 commitments and k=19, run on a server with 128 cores.
The proving time took 77 minutes. Out of it, only for 15 minutes was the core utilization at 100% and most of the time it stays at 15%.

It would be great to:

• Figure out if this cores under utilization only happens with a high number of cores
• Analyze which steps are being done in the 15% utilization periods, and see if we can parallelize them better.