Think about how to extend predictive models to PK models

[DavAug]

The logic of the PredictiveModel sampling, and visualisation with PredictiveModelPlot makes sense as long as no dosing information is required. But when we start to sample from PK and PKPD models it makes sense to

1. Be able to set the dosing regimen, either by hand or from a dataset. Potentially easiest way is to accept a pandas DataFrame in a standard format that contains dosing events.
2. When we sample from the predictive model, the dosing regimen information is returned, such that it may be visualised in a PKPredictivePlot.