When performing an attack step against trace index 0 we need to provide
the absolute pre-state rather than pulling the value from the middle
of the trace.

This PR also modifies the absolute pre-state that is stored in the devnet
to match what the op-challenger is playing.