Extend multiproof format to support bytecode & storage tries

[s1na]

The existing multiproof format can rebuild a trie from leaves to root. In Eth1 the accounts (located in leaves) store a codeHash and a root hash of their storage tries. If we want to send one multiproof for both accounts and their storage tries we'll need to add a few more opcodes to the proof format.

Alexey has apparently faced the same problem and has an extension of the original format, from which we can draw inspiration.

[gballet]

So after discussing this, I'm writing down what we intend to do on a first attempt. I will start to only implement an extra ACCOUNT instruction that takes whatever node is available on top of the tree, and push back an Account(balance, codeHash, stateRoot) node whose stateRoot member is the node that was popped from the tree.