Remove unproven shortcut from safeScalarMultiply

[jml]

Makes ensureInGroup behaviour conform better to Python implementation.

@ocheron care to take a look?

[ocheron]

Yes that should be enough to fix the issue with ensureInGroup.

However I'm wondering about any performance impact. If I understand correctly the derivation of scalar from password, we have 16 additional bytes here. So the scalar multiplication has to process 128 more bits.

It might be interesting to move the reduction modulo L to a place where the element is known to be in the subgroup:

scalarMultiply _ n x = safeScalarMultiply n x

[jml]

Thanks! Suggested optimization legit, but not something I have time to process right now—alas.