make device capable of stacking and concatenating usize type, which are necessary for preparing tensors for embedding (#882)

Co-authored-by: Corey Lowman <[email protected]>

Author: optman

dfdx-core/src/tensor_ops/utilities/device.rs

@@ -15,8 +15,11 @@ pub trait Device<E: Dtype>:
 
     // appends
     + super::super::stack::StackKernel<E>
+    + super::super::stack::StackKernel<usize>
     + super::super::concat::ConcatKernel<E>
+    + super::super::concat::ConcatKernel<usize>
     + super::super::concat_along::ConcatAlongKernel<E>
+    + super::super::concat_along::ConcatAlongKernel<usize>
 
     // optimizers
     + super::super::adam::AdamKernel<E>