feat(evm): use completely separated storage sections in multifork

[mattsse]

Motivation

with #1715 we added multifork support, with some assumptions/limitations:

• only one local memory area, which meant that write operations were performed in the same memory area regardless of forking

this has turned out to be an undesired and potential pitfall.

This rewrites how storage is tracked:

• each fork is completely separate, write operations are performed only in the fork's own separate memory
• the storage of the account of the sender and the contract remains persistent across swaps: e.g: changes to the test contract's own variables are persistent when selecting another fork

Solution

• store multiple forkId -> ForkDB pairs
• also keep track and update Subroutine when selecting forks
• drastically simplify the FuzzWrapper which is now a Cow<Backend>

[gakonst]

clean, anything else we should test beyond storage values?

cc @hexonaut, @aureliusbtc & @mds1 for any user testing we can get before merging in case we're introducing any edge cases here.

[hexonaut]

I'm getting a non-descript "revert" error in this test at this line with this PR (previously working): https://github.com/makerdao/xdomain/blob/add-optimism/packages/dss-bridge/src/tests/domains/optimism/OptimismIntegration.t.sol#L292

[mattsse]

@gakonst I think this is good to merge, but maybe we should wait until I updated all the docs, also need another round of revm debugging to figure out how to enable diagnostics for calls on non-existent errors.

But from a functionality point of view it is complete and can be merged

[gakonst]

Ack - reviewing today. Want to give some thinking on whether we can do this without the persist cheats

[mattsse]

@gakonst figured out the problem with diagnostics,

now we return meaningful revert reasons if someone calls a contract that only exists on another fork:

foundry/forge/src/multi_runner.rs

Lines 1210 to 1213 in df7ecd8

	assert_eq!(
	result.reason.unwrap(),
	"Contract 0xCe71065D4017F316EC606Fe4422e11eB2c47c246 does not exists on active fork with id `1`\n But exists on non active forks: `[0]`"
	);

[mattsse]

I'm going to revise all multifork docs next. Should we hold off on merging until tonight's night is built so we can ship it with tomorrow's nightly?

[gakonst]

Yep - let's give this one final pass and merge/ship tomorrow. I'd try to clarify to the reader how to think about all the forkign cheatcodes, which ones are the "most common" ones and which ones require deeper understanding.

[gakonst]

@hexonaut @mds1 merged, see how to use here. This should allow us to have contracts that persist across forks beyond the default test contract and msg.sender.