Add argument to learn() to condition on some variates

[pglpm]

In some studies one might want to draw nonparametric inferences on a subset of a dataset, determined by specific values of one or more variates – that is, calculating probabilities conditional on some particular variates.

At the moment this requires subsetting the dataset, removing the fixed variates from the metadata file, and using these modified files in inferpopulation().

As help to the user we could instead add some sort of conditionon or subset kind of argument in inferpopulation()? The conditioning would then be done internally.