chore(sol): further optimize solidity verifier (#16021)

## Overview 

Applies a few optimizations to the yul honk verifier cutting of 37242
gas

- Pre calculates the inversions for each round of sumcheck this saves
10k or so
- very aggressive loop unrolling saving 20k gas
- adds templating into cpp code to generate some of the unrolled
sections based on the LOG_N circuit size - mainly when used in the
inversions code
- unrolls the hot loop in sumcheck - using precomputed values where
possible

- reuses memory offsets for the sections of the code that calculate
inversions
    - barycentric domain inverses use scratch space
    - gemini inversions use the same scratch space 
- this memory is then overwritten by the vk when we accumulate in the
final steps

Author: Maddiaa0