Fix tests and/or provide test running documentation

[carlschroedl]

As mentioned in #14 and #16, I haven't been able to execute the tests successfully. To ensure continuous code quality, it would be awesome if we could accomplish some combination of:

• fixing the tests to resolve any errors
• fixing the implementation to resolve any errors
• updating the developer setup documentation (README.md)
• updating the continuous integration configuration (.travis.yml)

[r-barnes]

Most of the tests should probably live in compactnesslib, since that provides the underlying functionality to the library. We could either duplicate some of those tests here, or just target the interface between the two for testing.

[carlschroedl]

Cool! I get how we could duplicate tests. I don't understand the other option yet. Do you have any advice on how we would target the interface? Maybe a best practice blog post, or a link to an exemplary project? I haven't worked on any integration between Python and native extensions before, so I'm all about the learns :)

[r-barnes]

I'm not sure what the best way forward with this is either, other than that any critical testing of the internal calculations should probably be in compactnesslib rather than the wrapper.

If folks have test ideas, one option might be to write them in Python and I could then backport them to compactnesslib.

[carlschroedl]

Would it be valuable for us to duplicate a happy-path subset of the c++ lib's tests in Python? It would permit the benefit of some quality assurance on each pull request to python-mander without the burden of duplicating every test in the C++ lib. On the other hand, if python-mander is going to be a simple auto-generated thin wrapper around compactnesslib then those tests wouldn't bring much value because they'd really be testing whatever auto-generates the Python based on compactnesslib.

[carlschroedl]

Sorry, I just glanced up to your older post and saw that you are on board:

We could either duplicate some of those tests here