Add cofactors to curve interfaces

[AntoineRondelet]

For now, I can't see any use of cofactors in the current curve implementations (this may be because cofactor = 1 in bn128 etc).

That being said, I think it would be nice to add cofactors to the curve interface in order to pave the way for new curves support (e.g. bls12-{381, 377}).
In addition to that. It would be nice to add some explicit subgroup membership tests in functions like the deserialization (i.e. operator>> etc) to make sure we always process elements in the safe subgroup (where hardness assumptions hold).

Actually, isn't there a missing "multiply by cofactor" step in the edward curve implementation (cofactor = 4 there)?

[AntoineRondelet]

The associated PR was merged to staging. Closing this issue as a consequence.