add jackknife module

[goldingn]

i.e. refitting a model several times, each time dropping one covariate, then comparing the goodness of fit between these to see which covariate has the most importance.

This could be tackled in one of several ways:

1. a process module to drop a single covariate (this could be listed to explicitly do the jacknife)
2. a process module to do jackknifing for a specific modelling method
3. a model module to specifically do jackknifing for that modelling method
4. a process module to do jackknifing with a model module call passed as an argument (similarly to how TargetGroupBackground uses an occurrence module call)
5. something else that requires changing how zoon works

My preference is for 4, I think. Though it would be nice if zoon could handle that more easily