Short scalars in Shplemini: edge cases

[iakovenkos]

To make the UltraZK and MegaZK recursive verifiers (with UltraCircuitBuilder) more efficient, we use bn254_batch_mul with short scalars, at the moment, it's not handling the edgecases properly (def not handling repeated points). However, if the proof is generated with MegaZK or UltraZK Flavors, the probability of edge cases in a valid proof is negligible, because the witness commitments are masked. We need to

• Ensure that the collisions are also unlikely for the selectors.
• Investigate a robust way to process (whp malicious) proofs that contain edge cases.