Improvement of utilities for TwoLevelTree

[JulStraus]

This PR combines two topics:

1. bf23cec includes a default value for probability_branch and
2. eb97da6 updates utility functions (and their tests) as well as adds some.

A key factor for the changes to probability_branch is that we have to include it in practice in all multiplications for CAPEX contributions to the cost function. Otherwise, we would end up counting each branch as if the probability is 1. In this respect, the function is similar to probability. However, we could improve in the documentation that one should either use probability or probability_branch as the latter is included in all other TreePeriods or AbstractTreeStructure.

Regarding the utility functions, I would suggest we think whether we should export some of them. Specifically, I used n_strat_per in one of the changes I plan to implement within EnergyModelsBase.

[trulsf]

I am not sure I see the need for the get_strat_node function. Is it made just to be used in the tests?

I guess, some of the utilities can be exported. At the same time, I am bit concerned about exposing to much functionality for interacting directly with the tree structure. Is it not enough to be able to iterate through the nodes of the tree and the scenario paths to each leave? At least, I think we should have some good use cases where it seems relevant.

[JulStraus]

I am not sure I see the need for the get_strat_node function. Is it made just to be used in the tests?

Honestly speaking, I do not recall it right now. I probably lost a bit track of the different branches and stashes I have within all EMX packages ;) I think however it is good to have it for a latter stage (but without a docstring and internal). It is in general preferred to use the iterator utilities, but sometimes this is, at least in my view, not really possible as we do not consider all potential applications.

At the same time, I am bit concerned about exposing to much functionality for interacting directly with the tree structure.

I agree. I am currently using at least the functionality for identification of the numbers, i.e., n_strat_per, n_children and n_leaves. The question is whether they should be available or not.

[JulStraus]

The last two commits just update the exports and tests. The former was discussed above, the latter was essentially a requirement.