Opveclib

Upgrade to latest version of TensorFlow
Update CLA

This release includes the full opveclib optimizer which will automatically fuse allowable operators and their gradients at run-time to for use in a TensorFlow application.

It also includes improved documentation, additional API functionality and the use of integer types in input and output tensors.

First release of opveclib which includes the base functionality for defining TensorFlow operators. Release features include:
- Python-native operator definition which emits C++ and CUDA versions and links them into TensorFlow.
- Python-native gradient definition.
- Preliminary optimizer, allowing for operator fusion. Currently only activated for the OVL test framework (i.e. by using the evaluate() function).