set_refund_for_current_tx

[montekki]

What ❔

Fixes refund-related bugs in the glue code for the new VM:

• Computes pubdata_published as a difference with the previous transaction
• Sets refund for the bootloader state

[slowli]

Locally, the changes seem to allow the load test set up to complete, and the load test itself now runs fine 🎉 There are still issues with batch executor unit tests though, like extra storage logs and L2-to-L1 logs being produced by the new VM.

[slowli]

Please remove if this code isn't needed. BTW, how would pubdata computations work if the entire batch is executed (i.e., execution_mode == VmExecutionMode::Batch)? I'd think that pubdata_before needs to be updated somewhere in run() (in TxHasEnded hook handler?).

[montekki]

I've changed it to be updated here on the spot, as for now i do not have a better theory about where it should be updated.

[slowli]

I don't think it's correct (it would probably lead to the previous issue of pubdata_published for a transaction including pubdata produced by the bootloader / system contracts after executing the previous transaction), but let's fix this sequentially.