About candidate chosen for branching with a NaN score

[SelinBayramoglu]

I would like to ask a question that was also mentioned in #307

In question (2), a variable is selected for branching with a score of nan as a result of running

observation['scores'][0][action_set].argmax()

I am wondering if this is a bug because a variable is selected for which a branching score is not calculated.

Thank you!