perf: optimize Pedersen and IPA commitments gadgets

[yelhousni]

Description

This PR optimizes Pedersen and IPA commitments gadgets, through the use of joint_scalar_mul_be and scalar_mul_le implemented here arnaucube/ark-r1cs-std#1.

Benchmark

• The Pedersen check circuit goes from 3,575,519 r1cs to 2,455,364 r1cs.
• The IPA verify circuit also goes from 77,175 r1cs to 73,367 r1cs.

[arnaucube]

This is really great!

Over these past couple of months, Sonobe was updated to use arkworks v0.5.0, which slightly changes the types, but I think that will have not much affect to the changes of this PR, but it might affect more the fork of r1cs-std.

How would you see on doing the following?

• doing the PR of ark-r1cs-std repo (ark-r1cs-std PR) directly to the main arkworks repo https://github.com/arkworks-rs/r1cs-std
• and the current sonobe PR be done from latest main branch, where independently of the arkworks r1cs-std being merged to main, we can already use your fork with the r1cs-std improvement (of the arkworks-rs/r1cs-std (v0.5.0))

Alternatively we can merge the current PR (since it points to the branch decider-circuit-bench) and then update that branch to latest Sonobe main, while in parallel porting the r1cs-std changes to the last arkworks-rs/r1cs-std version, but probably the above will be cleaner.

In any case, happy to do any option that it's easier and less time consuming for you. Thank you very much for this improvement!!! 🙌