How does `max_features` limit the number of features in an sklearn ensemble model?

Question

I still do not entirely understand max_features in sklearn classifiers. The documentation leaves a little bit of room open for interpretation. For