Revisiting Cloud Deployment

[acannistra]

This is an effort to streamline the opportunities we present for others to use this software.

In particular, we hope to

• provide a Docker container that someone can use for pre-processing data in preparation for model training.

Some parts of this should be relatively straightforward, including a Jupyter Lab instance with the appropriate environment.

Others that are more complicated include:

• Re-structuring the repository to contain all "deployment-relevant" files and code in one directory (./deployment). This involves moving the Sagemaker Dockerfile and all relevant infrastructure (including Travis build instructions) around and updating paths.
• Rethinking the terraform deployment that we use to add all of this infrastructure to an individual's AWS account (this has needed revisiting for a while).

[acannistra]

(travis will fail until we update paths in master)