Review: 05: Chris Wood: Branching in Prio3 reduces readability

[cjpatton]

In particular, the sharding and preparation algorithms have lots of branching for generating joint randomness or not (depending on if JOINT_RAND_LEN == 0). Figure out how to refactor these algorithms so that the branching logic is handled by auxiliary functions.

[cjpatton]

The sharding algorithm is now much more clear. While technically we could add variants for preparation with and without joint randomness, I think the gains here would be less significant, as the code is already pretty short.

Closing this as completed.