Physical activity and mortality: is the association explained by genetic selection?