Set up experiment definition JSON files

[stevehadd]

Now that the experiment framework is in place, we need to create some experiment definition JSON files which can be used to run train, inference and evaluation in batch.

this should include:

• decision tree, input features=country,max_depth,year
• decision tree, input features=lat,lon,max_depth,year
• decision tree, input features=country,lat,lon,max_depth,year
• random forest, input features=country,lat,lon,max_depth,year
• nearest neighbour classifier, input features=country,max_depth,year
• neural network (MLP), input features=country,max_depth,year

The next step is to branch out into other ML libraries for classifiers. Further experiment definitions will then be defined for those libraries and classifiers.

[stevehadd]

There are now several experiment definitions, sufficient for the format to be settled. As we add more classifiers, libraries etc, more will be created, but I think there are sufficient for a first attempt to consider this issues resolved.