Use prysmaticlabs/hashtree to accelerate Merkle computations

[ralexstokes]

We can accelerate the computation of a parent layer of a Merkle tree using this library: https://github.com/prysmaticlabs/hashtree

The core routine for this computation currently is in this function:

ssz-rs/ssz-rs/src/merkleization/merkleize.rs

Line 96 in 84ef2b7

fn merkleize_chunks_with_virtual_padding(chunks: &[u8], leaf_count: usize) -> Result<Node, Error> {

It would be nice to add this library behind a Cargo feature flag, and write some benchmarks to determine if we get substantial speedups on realistic workloads (e.g. computing hash_tree_root on a mainnet BeaconState).

See #93 for a bit more context.