Automating Autograding: Large Language Models as Test Suite Generators for Introductory Programming