Model needs systematic way to determine the contact rate trend's t*

[karimn]

See #28 and #31

[wwiecek]

There are four issues here:

1) We'd prefer to define t as beginning, not the midpoint of decrease (when we vary the slope k, it's easier to reason about when something started, not when it was half-way through) 2) We could pick from a few definitions of t -- or make it into a parameter 3) We are not sure how hierarchical it is across countries. To complicate matters further, the logistic decrease can have many interpretations; e.g. it can be related to heterogeneous mixing (and then it would progress proportionally to epidemic) or change of behaviour (then it would be more invariant across countries or regions). 4) We need a way to compare different models to help us reason about what formulation of logistic decrease to use.