Add support for plotting surveillance algorithms

[dirkschumacher]

Often you want to have some kind of baseline data that tells you if the observed counts are higher than expected.
Numerous methods have been developed and many have been implemented in the surveillance package.
Adding a simple way to add a surveillance algorithm to the plot would be pretty nice.

A surveillance algorithm is usually fitted/trained on a set of previous data (e.g. a rolling time window of the last 7 days or maybe the last X years) and makes a prediction for the next unseen observation. Thus for each time point under surveillance the algorithm outputs an upperbound, expected value and a flag, if the observed count surpasses the upperbound. Some algorithms have a configuration parameters and some more output values, but the upperbound, expected value and flag for a set of time points is a pretty abstract interface for such an algorithm. Once could omit the expected value, but from my experience, the value is often requested by epis (n < 10 ;))

[zkamvar]

At first, I thought you were suggesting to add surveillance algorithms to incidence, but now I see that you are suggesting adding support for the output of these algorithms to the plots :)

I really like this idea! This would allow users to take advantage of the algorithms in surveillance and the plotting capabilities of incidence.