Fix sampling of guided MCDA methods

[Zapiano]

Currently there is one factor (:guided) responsible for flagging if that scenario should be counterfactual, unguided or guided, but this same factor is also responsible for indicating which MCDA to use in case of :guided scenarios. This way, if I want to run scenarios with :counterfactual, :unguided and :guided scenarios but, for the guided, use only one MCDA method, it involves a lot of gymnastics. Hence I think we need a way of choosing only one MCDA method.

I don't think that's possible with only the addition of a single method. This probably involves the addition of a new factor mcda_method.

[ConnectedSystems]

To address this cleanly, we need to incorporate support for categoricals

https://categoricalarrays.juliadata.org/dev/using/