An Approach to Guide the Search for Potentially Hazardous Scenarios for Autonomous Vehicle Safety Validation