Proper treatment of zero-knowledge

[ulrich-haboeck]

According to the recent writeup A note on adding zero-knowledge to STARKs we need to correct our randomization procedure. In particular, we need to

• add a masking polynomial in the batching step of FRI (the BSCR technique) to assure no information leakage from the folded oracles, and
• randomize the overall quotient component polynomials as described in the paper. Last but not least, we should
• double-check if the degree of added witness randomness takes into account extension field openings properly.