add spmm_coo op & kernel (as a part of GCN)

[yuang-chen]

SpMM COO

NOT ready for merging!

Add the basic spmm_coo to perform SpMM operations needed by GCN.

Only the CUDA kernel is offered.

Code pieces are borrowed from prior oneflow-GCN implementation.

GCN

GCN is implemented using the spmm_coo op, denoted as CooGCN. Compared with the PyGCN by Kipf, the performance of CooGCN running on Cora graph with 200 epochs is:

	CooGCN	PyGCN
Time	1.2032s	1.5967s
Loss	0.7060	0.7612
accuracy	0.8070	0.8150

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}