InteractionValues.get_n_order() is not safe for large player numbers.

[mmschlk]

The creation of a new InteractionValues object from InteractionValues.get_n_order() currently loops over a restricted powerst. While the powers is restricted it still can be unsafe for larger number of players. Then often a lot of values are anyways zero and would not need to be stored:

max_order = order if max_order is None else max_order
min_order = order if min_order is None else min_order

new_values = np.zeros(
    count_interactions(n=self.n_players, max_order=max_order, min_order=min_order),
    dtype=float,
)
new_interaction_lookup = {}
for i, interaction in enumerate(
    powerset(range(self.n_players), min_size=min_order, max_size=max_order)
):
    new_values[i] = self[interaction]
    new_interaction_lookup[interaction] = len(new_interaction_lookup)

We should iterate over the InteractionValues.interaction_lookup dict instead.

This should be adapted and checked with a unit test that it works for a very large players and a sparse InteractionValues object.