Refactor ProblemModellingController for population inference

[DavAug]

Use case: Simultaneous inference on control group and dose group.

• Automatically reduce outputs of model when not all outputs where measured.
• Add optional field to HierarchicalLogLikelihood which can be used to map the parameters to the population models
• Construct masks for the parameters of the individual log-likelihoods and masks for the inputs to the population models.
• Masks for individuals most efficient if array of indices are constructed
• population models can be evaluated using slices according to the ordering
• The parameters are ordered as before but some individual parameters can now be dropped if they don't exist on the bottom level.