Rework CommitmentScheme abstraction

[swasilyev]

Current trait requires self to be used for committing/opening and verification at the same time.

/// Polynomial commitment scheme.
pub trait CommitmentScheme<F: PrimeField, P> {
    type G: AdditiveCommitment<F>;
    fn commit(&self, poly: &P) -> Self::G;
    fn commit_const(&self, c: &F) -> Self::G;
    /// Verifies that `p(x) = y` given the commitment to `p` and a proof.
    fn verify(&self, commitment: &Self::G, x: &F, y: F, proof: &Self::G) -> bool; // TODO: generalize proof type
}

These types has to be separated as they may be very different pieces of data. For KZG verifier key is 1G1 + 2G2 elements, while committer key can be huge.

[swasilyev]

Reworked in #8 Good enough so far. Makes sense to have a non-trivial implementation, before polishing further: #9