[Bug]DGL sparse run error message is incomplete

[paoxiaode]

🐛 Bug

DGL sparse run error message is incomplete, make it hard to debug for user.

To Reproduce

In DGL sparse, the op like bsddmm needs all inputs are on the same device, howerver, if inputs are not on the same device, it won't report "they are not on the same device" but report Segmentation fault, it is hard to debug for user.

Steps to reproduce the behavior:

>>> import torch
>>> import dgl.sparse as dglsp
>>> indices = torch.tensor([[1, 1, 2], [2, 3, 3]])
>>> val = torch.arange(1, 4).float()
>>> A = dglsp.spmatrix(indices, val, (3, 4))
>>> X1 = torch.arange(0, 3 * 5 * 2).view(3, 5, 2).float() 
>>> X2 = torch.arange(0, 5 * 4 * 2).view(5, 4, 2).float()
>>> X1 = X1.to("cuda:0")
>>> X2 = X2.to("cuda:0")
>>> dglsp.bsddmm(A, X1, X2) // A is still on the CPU
Segmentation fault (core dumped)

Expected behavior

Should we add the input device check for the DGL sparse operator?