Efficient Learning-Based Sound Propagation for Virtual and Real-World Audio Processing Applications