On the rotation and frictional lock-up of normal faults: Explaining the dip distribution of normal fault earthquakes and resolving the low-angle normal fault paradox