Covariate Balancing Methods for Randomized Controlled Trials Are Not Adversarially Robust