For the analysis of imperfect recall (and also absent mindedness, see #578), it is useful to know, for a given information set, which are the most recent action(s) of the same player which precede this information set. For example, this is always a singleton in the case of perfect recall.

Computing these is already implemented internally. We will expose these as GameRep::GetOwnPriorActions(GameInfoset) in C++ and Game.get_own_prior_actions(infoset) in Python.

And as always add appropriate tests to exercise this. Because these data relate closely to testing imperfect call, tests of this function may help streamline and focus imperfect recall tests.