docs: update `VartimeRistrettoPrecomputation` documentation

[AaronFeickert]

It is the case that performing variable-time Ristretto multiscalar multiplication evaluation using VartimeRistrettoPrecomputation may be less efficient than the corresponding VartimeMultiscalarMul implementation, depending on the number of points used.

This PR updates the documentation to make this clear, since the decision of whether or not to use precomputation may depend highly on the use case.