drop in replacements for GP to estimate infection timeseries

[smwindecker]

We should consider a suite of underlying models for infection incidence timeseries, including more mechanistic ones, for future extensions of the software.

drop in replacements for GP

[AugustHao]

in those implementations we might need a time-varying, flexible deviance term (eg a GP over time) to absorb the residual deviance between the mechanistic model output and the observed data, much like component 2 in the TP reff model

[smwindecker]

enable user to account for susceptible depletion in reff forecast.