Generalize Accumulator, VectorCommitment over types that implement Hash

[eddiew]

For example, we can generalize
Accumulator<G: UnknownOrderGroup>
to
Accumulator<T: Hash, G: UnknownOrderGroup>

and add goes from
fn add(self, elems: &[Integer]) -> (Self, MembershipProof<G>)
to
fn add(self, elems: &[T]) -> (Self, MembershipProof<T, G>)

The type parameter T in the accumulator struct would be a phantom type to help prevent mixing accumulators for different element types

[whaatt]

I think this is overall a good idea -- one thing to consider though is that if hash_to_prime is a bottleneck, people might want to cache its Integer results for use with the accumulator functions.

Granted, hard to say if people will be doing that in practice...