fix: avoid creating extra journal entries

[klkvr]

Motivation

Closes #7481

The actual reason behind the code in issue failing is that there is an underflow happening after one of vm.transact calls. However, we should have just printed traces ending with [Revert] panic: arithmetic underflow or overflow (0x11) here.

The panic happens because current impl of apply_state_changeset updates state entries by doing state.remove() + load_account, which results in state.journal having JournalEntry::AccountLoaded item for each account in changeset.

Then, when call to test contract reverts due to underflow, revm firstly goes through AccountLoaded entires and removes those accounts from state: https://github.com/bluealloy/revm/blob/main/crates/revm/src/journaled_state.rs#L318

Because of that, on the next, for example BalanceTransfer entry, it will assume that account is still in the state, but it was just removed earlier and a panic will happen: https://github.com/bluealloy/revm/blob/main/crates/revm/src/journaled_state.rs#L351

Solution

Reuse update_state fn used by --isolate which updates state without touching journal

[klkvr]

Just realized that what vm.transact is exactly what we are doing with isolation mode, it might make sense to merge logic for those at some point

[mattsse]

nice find, one pedantic nit re error handling

you're right this is basically --isolate, so it would be beneficial if we could unify all of this eventually

[mattsse]

test failures unrelated, but we have some new unrelated clippy issues, mind taking care of them @klkvr ?

also I think it's time to retire the goerli fork tests