Design Sensitivity and Its Implications for Weighted Observational Studies