Broken tests

[burtonjosh]

Current tests are broken, needs fixing

[burtonjosh]

One test comparing the output from an old python implementation (which is arguably not super reliable) is quite broken (marked with @test_broken in #16). The only change I can think of that would have caused this is updates to the solvers or MethodOfSteps.

Would be nicer to have some lower level tests checking functionality of the filter at each step. That would help with future debugging, since now it's not clear what goes wrong when a change is made.