Hash the account trie along with storage tries in parallel

[rjl493456442]

It has been observed that computing the account trie hash takes a few milliseconds, which is non-trivial in the context of block execution.

The idea is to parallelize account trie hashing with storage trie hashing by including all available accounts in the hashing step once their corresponding storage tries have been committed.

Note: the necessary trie nodes for account hashing are preloaded in advance, no significant I/O overhead is expected during this process.

[jwasinger]

Assigning myself as I've been working on a PR for this.