Improve testing of long-running PHYX files

[gaurav]

There are some PHYX files that take too long to test on Travis. My plan for dealing with those is as follows:

1. PHYX files that can be tested quickly are kept in a phyx directory and are automatically and continuously tested using Travis CI. We need to ensure that these files are representative of the different kinds of phyloreferences we have. We should try to keep the time taken to run all these files to within 60 minutes (we're currently at around 30 minutes).
2. PHYX files that take longer to run are kept in a phyx_too_slow directory. There should be a mark (i.e. a test command line option, such as py.test tests/ -m runslow) that tests all the slow-running phyloreferences.
3. For ad hoc testing, phyloreferences can be reasoned over in Protege and then exported as an OWL ontology to be used with jphyloref's --no-reasoner. We should also create a script that tests a particular PHYX file.
4. All these changes should be documented in the README.md file when implemented.

Our long-term goal is to speed up reasoning so that more files can be included in the Travis run, but this should be good enough for the short-to-medium term (i.e. until after TDWG 2018).

[gaurav]

Another possibility: we could store a Gzipped file of the reasoned ontology and run that in the test suite. Since reasoning is the slowest part of the process, this should run pretty quickly. The one long-running test file we have is around 800kb as a Gzipped RDF/XML file, well under Github's 50MB limit; we might be able to make it smaller by using Turtle or another export format.