Set up code structure for initial modeling pipeline

[LswaN58]

Need some sort of pipeline structure. I think we'll want another manager/processor, a generator subclass, and probably a registry. For now, entire implementation can happen in the generator, and we can separate out filtering stuff further down the line when we get the filtering classes in common.

But for placeholders, we should have functions for setting filters on different features (or columns) for training phase. Also, for our calculation step, we should actually have two functions: Train and Apply.

[LswaN58]

Concretely, doing the following:

• ModelManager - like other Manager classes, this will have functions to call its generators to process each event/featuredata. It will also need to keep track of individual Processor instances - for now, this is just a single population-level one.
• ModelProcessor - this probably ends up like PopulationProcessor, where there's only ever one instance, but we do need the functions to distribute events and/or features to generator instances via a registry.
• ModelRegistry - Probably 90% the same as an ExtractorRegistry, just needs to load up "Models" instead.
• Model - Generator subclass, can 'extract' from event or feature data, which probably just looks like tacking them onto lists. Not very efficient, realistically we should just have our "Train" function directly operate on a given array. Not how we're set up now, though, so it'll have to come later. Then instead of just a CalculateFeatureValues or GenerateEvent or whatever, we'll have two "calculation rule" functions: Train and Apply. Train runs when all events are in, generates actual model object.