Optimize the polynomial interpolation in lib.rs.

[afck]

#23 improved polynomial interpolation a lot, but there is another implementation in lib.rs, which, instead of computing the coefficients, just computes the value at 0. It can probably be improved in a similar way. For this, Neville's algorithm is supposed to be faster.

We should benchmark this function (which is used in threshold signatures and decryption), too. (See #16.)

[afck]

Neville's algorithm is considerably slower, probably because it performs more arithmetic operations in the curve as opposed to the base field.