<!--- Provide a general summary of your changes in the Title above -->
<!--- Describe your changes in detail -->
This pull request implements the grt123 identification algorithm.
It is based on PyTorch, which currently conflicts with Tensorflow, so the pull request also disables the Keras import.
While the algorithm is quite fast on GPU, it is painfully slow (but working) on the CPU due to a bug in PyTorch. The bug is fixed in the PyTorch development branch, but the next release is about 2 months off.
## Reference to official issue
<!--- If fixing a bug, there should be an existing issue describing it with steps to reproduce -->
<!--- Please link to the issue here: -->
## Motivation and Context
<!--- Why is this change required? What problem does it solve? -->
<!--- If adding a new feature or making improvements not already reflected in an official issue, please reference the relevant sections of the design doc -->
This Pull request implements part of the prediction service (https://concept-to-clinic.readthedocs.io/en/latest/design-doc.html#detect-prediction-service)
## How Has This Been Tested?
<!--- Please describe in detail how you tested your changes. -->
<!--- Include details of your testing environment, and the tests you ran to -->
<!--- see how your change affects other areas of the code, etc. -->
The algorithm has largely been tested by manually confirming about 5 random LUNA2016 datasets - as the loaded model has already been tested extensively elsewhere, this was deemed sufficient.
It has been tested both outside of docker, on a Titan X GPU, and inside docker on an i7 Intel CPU.
## Screenshots (if appropriate):
- [X] I have signed the CLA; if other committers are in the commit history, they have signed the CLA as well