Translation evaluation correctness in the presence of ZK

[maramihali]

For zero knowledge, we use random scalars to hide the witness polynomials. This needs to be done in a way that still allows us to do the translation evaluation check for the VMs.

If we establish that random scalars generated from BN254 are still uniformly random on Grumpkin, we can use the same randomness to hide the ecc_op polynomials in ECCVM and Translator and the evaluation check remains unchanged. Otherwise, Sergei has come up with a subprotocol.
https://hackmd.io/p1VE4F3fT5OKBW7muJX28w?view