Read contract bytecode from latest even when execution block is mined-past

[carneiro-cw]

Need to think a bit about the design of this, if naively implemented this means that a contract could be executed in a point in time before it was deployed, but I don't think that's so bad at all. If it is bad, then we'd need to track the deployment block of bytecodes in the cache, which might be hard in the current storage layout.

Worth noting that the contract code is immutable, so there is no risk of reading code in the wrong version.

[carneiro-cw]

I don't think the evm differentiates between bytecode reads and general account reads, so we need to identify some other way that it is requesting the bytecode, we can probably assume that the first account read of the "to" field is the call to get the bytecode or something.

For multicall this would load the contract from the cache, but then when multicall calls the actual contracts they would still be read from perm.

We COULD however have special input parsing for multicall to know which calls it is making (useful for logs and metrics too), will open a separate issue for this.