Identification of non-minimal adjustment sets

[AndrewC19]

Our current identification algorithm only enables the user to automatically identify minimal adjustment sets (i.e. the smallest set of variables that blocks all back-door paths).

While this is usually preferable (less adjustments generally means a simpler statistical model), there will be some cases where a user wants to use a specific non-minimal adjusment set.

To this end, we should add a function to the CausalDAG class that obtains a list/set of all adjustment sets.

[jmafoster1]

I've just checked the code and it looks like the identification method just calculates the minimal adjustment sets. We can certainly include these as part of the test result, but I don't know how to calculate all the non-minimal ones. Dagitty has an algorithm for it, but porting it from R/JavaScript to python will be nontrivial.