Refactor Ristretto Point operations.

[CPerezz]

Reviewing the Ristretto implementation, and working also on #78 and #76 at the same time, I realized that the implementation done for Ristretto Point ops was not correct because:

The operations are mapped to the underlying EdwardsPoint that the RistrettoPoint represents (A RistrettoPoint is a wrapper type over EdwardsPoint).

This is correct, if you're working with an a and d values that satisfy the Ristretto constraints/requirements (a-d) is Quadratic Residue. But in our case, since we chose the @ebfull tweet parameters which do not satisfy the Ristretto req. we also use another d value for our RistrettoPoint which is different from the one used on EdwardsPoint formulas.

This causes:

• Encoding/Decoding functions are mapped to RISTRETTO_D instead of EDWARDS_D. (That's how it is implemented).
• Point op. functions have to be mapped. Since we are working with an isomorphic twist that already provides the same Scalar field, and Sub-group orders, we need to map the operations since the RistrettoPoint operations cannot rely on EdwardsPoint ones because they are already mapped to a different d.

That second thing was not considered at the beginning and so we need to refactor it.

[CPerezz]

With e935456:

• Doubling is the only operation that has been both implemented and tested.

• Mul is implemented with the double_and_add algorithm. It relies on addition and doubling, so due to the first dependency, it cannot be tested yet.

• Addition has not been implemented yet since we need to figure out the mapping between the isomorphic twist addition and the Doppio addition (both in TwEds Extended Coords)

[CPerezz]

Tests implemented in 252f782 proved that Addp and Mul impl are working as expected.