Not a modern one, but the 'cut-away' we used in groundschool had the axis across the case, with the rotor spinning forwards at the top. This ensures that the precession caused by the turn will tend to keep the rotor vertical (or at least closer to vertical) so that the yaw rate will be measured about the vertical to the horizon, not the normal axis of the aircraft.
Now if you want a really arcane discussion, let's talk about the turning errors of the artificial horizon!?!