Before training on labelled data, it's worth checking whether one label swamps all the others.
Task: write majority_label(labels) that returns the label that makes up more than half of the list. If no label does, return None.
None."cat" and "Cat" are different).