So we have two opposing schools of thought:
1) The very well informed aerodynamicist friend likes to have the CP moved inboard, meaning outboard rotation, to reduce rolling moment presumably. I won't argue his logic, except to point out the following -
2) Every conventional twin I'm aware of has its critical engine defined (or at least described) by P-factor theory; and when Piper chose to create the PA-39, and thus reduce Vmca, it rotated the props inboard to minimize P-factor.
In the absence of empirical tests disproving #2, I'm inclined to go that direction.