apply_mask filter: improve tests #139
Description
Adapted from #136 (comment):
At the moment I'm heavily leaning on parameterize to cover the range of tests, and then essentially implementing a parallel implementation - which I don't like at all. The options I can see are either to a) hard-code what the masks should look like (probably the best solution), but probably reduce the number of test cases (e.g. testing rename separately from masking); b) mock out the masking so we just check that e.g. np.less is called; or c) leave it as it is 😄
Probably some combination of a) and b) are best from a "good design" perspective
Ideally we should separate out concerns in these tests and move away from the "parallel implementation" approach