On the Computation of Size-Correct Power-Directed Tests with Null Hypotheses Characterized by Inequalities