How big, how bad, how often: are extreme events accounted for in modern seismic hazard analyses?