Add documentation for data formats

[ngreenwald]

Would be nice to add information about the file/folder structure the different models expect. For example:

• specifying that multiple files should be zipped together
• specifying that segmentation and tracking take in single-channel 2D data
• specifying that multiplex takes in multichannel 2D data, with nuclear channel first and membrane channel second
• specifying that multiplex only accepts multichannel tifs with channels last, which is not the typical image format

[willgraf]

While some information is added in #128, it would be best to include the required information on the /predict page to better inform users when they are about to select a model.

From #128:

It would be nice if there was some way to have the description of required input data for each model more easily visible from the model dropdown section on the website.

Right now, if someone goes to deepcell.org/predict, it wouldn't be obvious how to find out what the requirements are for each model, or even what the different models do.