Currently we are calculating the state root recreating the merkle tree from scratch each time we require it. (ie. here and here).

We should keep the world state in memory and update only the state transitions.

Its a follow up on #45 and there is a TODO comment in code