Small errors in endo mul

[95DDB]

A couple of errors with the endo mul must be fixed:

1. Tests (both for primitive and gadget) for Tweedle are wrong, as they consider the first 128 bits of the scalar, while they should consider the last (first are all 0s) (already addressed as part of 344d45f )
2. Like we did for mul_bits and mul_bits_fixed_base, would be nice to add a function that checks the inputs and returns error if they are not valid (e.g. trivial base point)
3. Let's change the interfaces to not accept Vec<bool> but Iterator<bool> (avoid unnecessary cloning)