Stwo logup-gkr, proving a single lookup

[ShuangWu121]

This PR adds logup-gkr support for stwo backend.

Only supports a single lookup operation for now.

more descriptions will be added later.

[leonardoalt]

@georgwiese @Schaeff @ShuangWu121 do we still want to merge this?

[Schaeff]

@ShuangWu121 could you please explain the state of this PR? What is supported and what isn't? Maybe updating the description.

[ShuangWu121]

@ShuangWu121 could you please explain the state of this PR? What is supported and what isn't? Maybe updating the description.

here #2543 I wrote more descriptions.
The GKR method is able to prove single lookup and multi-lookup without dynamic machine size now. But I did some initial tests of GKR performance, it is much slower than none-GKR when stwo uses Blake2 as hash function, it is better when stwo use Poseidon, the same as mentioned by Miden VM Logup-gkr implementation benchmark result. but I didn't do a vigorous performance test, as now the logup-GKR use BusInteration, but pil still generate bus constraints, stage 1 witness, accumulators etc. and in addition, it proves BusInteration using logup-GKR. To complete this PR, it would need generating BusInteraction Identity without pil generating stage 1 accumulators, thus needs PR #2629 #2631 #2636 and maybe witgen (I remembered @qwang98 mentioned witgen might not sound without accumulators ...).

Getting it to work with dynamic machine sizes will require more effort, and the final performance is still uncertain. So this PR might be paused for now, since we're currently focusing on pre-compiles?

[leonardoalt]

Closing for now, branch is still alive if we want to re-open.