Eth API: Trace installed bytecode in the Ethereum JSON-RPC trace_block output

[Stebalien]

The Ethereum JSON-RPC trace_block API currently returns 0xfe as the EVM bytecode for all newly created actors:

1. For non-EVM actors, this makes sense as there's way to return any reasonable bytecode.
2. For EVM actors, there's just no reliable way to get the bytecode because an EVM smart contract's bytecode can change multiple times over the course of the block, even without reverts, by invoking selfdestruct and re-deploying. Even EIP-6780 will maintain this behavior.

To fix this, we need some way to get the bytecode of an EVM smart contract in the middle of a block. We'll need to implement both:

• Expose intermediate state ref-fvm#1994
• Trace final actor state ref-fvm#1995

Then, every time we see a "create" event in the trace, we can lookup the state of the smart contract after it finished running the constructor to retrieve the bytecode.

[aarshkshah1992]

@rjan90 @jennijuju Why has this been prioritised ?

[Stebalien]

An interim solution is to only handle smart contracts that live beyond the end of the block. Basically:

1. While we process the trace, for each address (ID) at which a smart contract is deployed, record the last transaction that successfully (not reverted) created a smart contract there.
2. After processing the trace, for each of those addresses, lookup the associated smart contract.
3. If the smart contract has self-destructed and/or doesn't exist, ignore it.
4. If the smart contract exists and is alive, get its bytecode and replace the "fake" bytecode we recorded in the trace with the real bytecode.

Note: we're only altering the trace of the last successful attempt to deploy a contract at any given address. Given that we won't have any intermediate and/or reverted state, that's all we can do. On the other hand, that's probably sufficient in most cases.

[Stebalien]

Basically:

1. If a call and/or sub-call hasn't been reverted, pass a non-nil deployedCode map with a type map[ActorID]*EthBytes through the tracing logic.
2. When processing a contract creation and deployedCode is non-nil, record deployedCode[contractID] = &crateTraceAction.Code even if there's already an entry for contractID in this mapping.
3. After processing the trace, for each contractID, codePtr in deployedCode, lookup the associated contract, assert that it isn't dead, and set *codePtr = contract.GetBytecode().