Refactor linked lists initial hashing

[4l0n50]

This PR changes the logic for hashing the initial tries using linked lists. Instead of traversing the initial state trie and setting all the payloads, it inserts all the nodes into the initial linked lists state trie. This change saves around 10% 15% of CPU cycles. Even though #467 considered also refactoring the final hashing, it turned out that the current implementation of the final hash is more efficient.
Closes #467.

[hratoanina]

Looks good!

Just a general question about soundness not specific to this PR:
For the initial trie, we check consistency of the state MPT by adding all the initial accounts of the linked list in order. Since we don't use the next pointers of the items and just advance by @ACCOUNTS_LINKED_LISTS_NODE_SIZE, what happens if a malicious prover provides duplicate accounts? I can't see a way to abuse it but I'm not sure it's completely safe either.

[4l0n50]

For the initial trie, we check consistency of the state MPT by adding all the initial accounts of the linked list in order. Since we don't use the next pointers of the items and just advance by @ACCOUNTS_LINKED_LISTS_NODE_SIZE, what happens if a malicious prover provides duplicate accounts? I can't see a way to abuse it but I'm not sure it's completely safe either.

Good point! I think this is indeed a problem. Even though the only account in the trie would be the last one, It seems possible to have multiple accounts with different states (e.g. different balance, storage trie) and choose any of them when writing/reading from the state. This might certainly allow you to take the last account from the duplicates to an incorrect state, getting a corrupted final hash.
I think the solution is to add the check addr_key < next_addr_key so that you never read/write from the duplicated accounts/slot.

[4l0n50]

@hratoanina I addressed the soundness issue we discussed. Now it's checking well formedness of the initial next node ptrs, as well as checking that the initial keys are strictly increasing.