Interval Estimation of Individual-Level Causal Effects Under Unobserved Confounding