Confronting the Reproducibility Crisis: A Case Study in Validating Certified Robustness