Optimization-based Structural Pruning for Large Language Models without Back-Propagation