Add a test that cannot be fooled by returning zero

Right now all of our tests are checking that something that should be zero is zero. This, however, can be fooled by making all or some of our outputs zero, and therefore is risky. An additional and potentially more reliable way to test the algorithm, for example, is to check that \operatorname{Tr} \tilde{\rho} O = \operatorname{Tr} \rho \tilde{O}.

EDIT: This test will also be fooled if both \tilde{\rho} and \tilde{O} are zero. Not sure what a good test would be.

If we have an example of something where we know the explicit answer, that'd be nice too.