Doubly Robust Fusion of Many Treatments for Policy Learning