The frequentist implications of optional stopping on Bayesian hypothesis tests