feat: custom gate for nonnative multiplication

[ivokub]

The checks performed in #749 are well-shaped for being implemented using custom gates and could give significant performance improvement.

As the initial target would be zkevm, then it would also be OK if we only implement the constraint system without the prover/verifier logic.

[ivokub]

I tried the approach very slightly - I'm not really sure it is worth it as we would have to add additional wires to make sense, add additional copy constraints etc. I'm afraid that the added overhead cancels out the efficiency of having less gates.