Security Analysis of Safe and Seldonian Reinforcement Learning Algorithms