Using Response Times for Modeling Missing Responses in Large-Scale Assessments