The ConceptARC Benchmark: Evaluating Understanding and Generalization in the ARC Domain