Dglke

This patch release provides the following features.
* Offline inference: https://github.com/awslabs/dgl-ke/pull/103
* Reorganize documentation.
* Fix bugs:
* 85 , Force user to provide dataset name when using udd or raw_udd. (This avoid set 'FB15k' as dataset name of user defined data.
* Minor improvements:
* remove unnecessary eval log: https://github.com/awslabs/dgl-ke/pull/104
* Add check for udd input
* allow users to specify the delimiter

We are happy to announce the first release of DGL-KE, a lightning-speed package for learning knowledge graph embeddings. DGL-KE was previously incubated under the DGL repository. It is now a standalone package with more efficient and more scalable training. The key highlights are:

* Effortlessly generate knowledge graph embedding with one line of code.
* Support for giant graphs with millions of nodes and billions of edges for various hardware:
* multi-CPU machines,
* multi-GPU machines,
* a cluster of machines.
* Support for both Pytorch and MXNet.