Benchmarking Heterogeneous Treatment Effect Models through the Lens of Interpretability