feat(trie): implement change set walker

[RomanHodulak]

About

This PR is about decoupling trait and implementation of an iterator over change sets in a transaction.

Motivation

Expanding on #9282 with a focus on state module of reth-trie crate.

Goals

• Remove dependencies on reth-db and reth-db-api in state module of reth-trie crate.

Solution

• Decouple trait and a database implementation of an iterator over a range of change sets of accounts and storage.
  • Create a new trait TrieRangeWalker and TrieRangeWalkerFactory
  • Implement above mentioned traits using a newly created DatabaseAccountChangeSetsCursor and DatabaseStorageChangeSetsCursor
  • Update all usages in reth-trie to use the trait only
  • Update all usages outside of reth-trie to supply the proper implementation

Additional notes

The newly created database implementations of TrieRangeWalker and TrieRangeWalkerFactory should be a part of reth-trie-db crate instead of reth-trie as done here. However, this is not a concern of this PR, instead will be a concern of a PR focusing on moving all database implementations of trie readers/writers to reth-trie-db. Mostly in favor of keeping the PRs small and focused as explained in #9585 motivation section.

[mattsse]

nice, this is becoming more clear to me
lgtm

pending @rkrasiuk

[RomanHodulak]

Hey @mattsse it's good to hear from you again, thanks for the quick feedback. I just got back from a week long break so I hope this effort didn't seem too abandoned!

Anywho, I'm glad that this looks more clear to you as I see clarity as important.

[mattsse]

that's great to hear,
we're getting a lot of requests for sp1 support which requires that some crates that currently depend on reth-trie types need to compile to wasm32-wasip1 and this is now blocker and slightly higher prio.
if you have the capacity to continue with this that'd be amazing

[shekhirin]

LGTM, minor nits

[rkrasiuk]

@RomanHodulak thanks for the PR, however I do not see any good reason for abstracting away the transaction from PrefixSetLoader given that it's only ever used with database and used for incremental root (the concept that now exists only in reth-trie-db crate), i'll open a quick PR that moves the loader to reth-trie-db crate

[rkrasiuk]

@RomanHodulak jumped the gun and misread the changes

[rkrasiuk]

@RomanHodulak i see, this is for HashedPostState::from_reverts. Still no need to abstract it over factory, just similar to other traits please create the trait in reth-trie-db and implement this method there

pub trait DatabaseHashedState<'a, TX> {
    fn from_reverts(tx: &'a TX, from: BlockNumber) -> Result<Self, DatabaseError>; 
}

[rkrasiuk]

closed in favor of #9987