A Survey of Constraint Formulations in Safe Reinforcement Learning