Rethinking and Accelerating Graph Condensation: A Training-Free Approach with Class Partition