Add trie diff tool for evm_arithmetization's CPU

[Nashtare]

Many of the errors we encountered on the CPU side where invalid final tries (usually the state trie). We currently print to the logs the tries computed by the CPU upon kernel failure, though this isn't really helpful on its own.

We already have a tool (eth-trie-tools) to highlight trie discrepancies between two (expected) identical tries which happened to have at least one diff point, but the process is slow and tedious. Ideally, we would like to have an automated tool that:

• keeps the initial / final tries generated by the decoder for a given payload (we normally hash the final tries)
• replay the failing payload on the evm_arithmetization side, and extract the resulting tries (see output_debug_tries() on how they are retrieved)
• look for a diff between expected final tries of trace_decoder and obtained tries of evm_arithmetization
  • for discrepancies in txn or receipt tries, display the distinct, decoded payloads
  • for discrepancies in the state trie:
    • highlight the account that's different
    • look for the associated address
    • if the issue lies in the storage root, find the discrepancy in the account's storage trie

This will be particularly helpful when we start working on actual blocks with SMT.

[BGluth]

Awesome idea! I don't know why I didn't think about calling the diff logic directly on failure.

[Nashtare]

Hey @atanmarko, just bumping this to know the status, if this is in your plans soon or not.

[atanmarko]

@Nashtare Plan to work on it during next week