feat: change oob behavior of RETURNDATALOAD and RETURNDATACOPY

[charles-cooper]

use zero-padding instead of exceptional halt

[chfast]

Do I understand it correctly that Vyper is in favor of modifying the existing opcode instead of replacing it with a new opcode?

[charles-cooper]

Do I understand it correctly that Vyper is in favor of modifying the existing opcode instead of replacing it with a new opcode?

Mildly in favor. It's not a huge issue, it just seems easier than adding a new opcode

(Fwiw my understanding is solidity is in favor too, but I can't speak for them)

[ekpyron]

To confirm: yes, we're also in favour of this - this significantly simplifies optimizations since with this returndatacopy can be safely removed in all cases without changing semantics.
Also, for us changing the behaviour of the existing opcode is easier than any other solution (like removing it and introducing a new one) and we don't foresee any problems with it.