Doubly Robust Inference on Causal Derivative Effects for Continuous Treatments