Add the ability to input the distribution details

[micdoher]

Which in turn populates the underlying yaml that deploys the corresponding pods. Adding it to the GUI means the data scientists (or pre sales) don't need to know about the K8s config aspects of distributed training