Add const quantization for torch.gather ("params" attribute only).#1227