running_sum appends each raw element instead of the cumulative total. Both suites require running_sum([1, 2, 3]) == [1, 3, 6]. After the refactor that comparison runs when the test requests the fixture, not in the test body.
