Training/Prediction Demo

We want to be able to demo the plugin at a friday meeting, either on 5/17 or 5/24.
@benjijamorris has already fine-tuned some parameters for the variance dataset, and has been able to achieve good segmentation results through the API.

We need to be able prove we can replicate that through the plugin on a windows workstation. This is what we need for the demo:

1. Models trained through the plugin that are able to give good segmentation results. We need good models for multiple (2+) variance datasets.
2. demo of models doing prediction, either through screen captured video or live.
3. parameters used to train the models, for reproducibility.

We have a windows workstation ready for @amilworks, and you could start by testing version `0.1.1.dev0` on artifactory (available w/ `pip install --index-url https://artifactory.corp.alleninstitute.org/artifactory/api/pypi/pypi-virtual/simple allencell-ml-segmenter==0.1.1.dev0` while connected to vpn). We'll release another version later in the week with user-defined patch sizes which we can install when ready.