Proper stdlib edge cases handling

[iakovenkos]

• Currently, we are using an insecure pattern in check_evaluations_consistency method. The problem in SmallSubgroupIPA is that an evaluation challenge may end up in the chosen multiplicative subgroup. Need to check in-circuit, whether r^|H| - 1 == 0.

• There's a similar issue in the SumcheckVerifier class in the stdlib context. We don't support barycentric evaluations at the domain points. Need to add logic for it to not cause panic.