Information leak when reconstructing wraps

[gmuraru]

Description

There is a possible information leak when doing public division. At one point we reconstruct the wraps and then we do something like:

value = value - wraps * 4 * (ring_size / 4)

Expected Behavior

The wraps should remain as shares and not be in plaintext

[nph4rd]

I'm taking this on... :)

[nph4rd]

I think it would be a good idea to add direct tests for this part of the code-base. Currently, we only cover it indirectly in the MPC tensor tests. This will allow us to have a better handle of the situation, I think. What is your opinion on this?