Towards Assessment of Randomized Smoothing Mechanisms for Certifying Adversarial Robustness