Improve testing

[pseewald]

trafficstars

There are two different testing mechanisms in place:

1. unit tests (hard-coded Fortran snippets)
2. testing by cloning Fortran projects

For 1) non-default command line options can be tested, but because fprettify is called as an external scripts, there are problems with python versions and test coverage is not reported For 2) we can test only with default options but test coverage works

Both mechanisms should be unified and both should call fprettify.run(). Then test coverage and non-default options should work.

Furthermore the testing procedure should be documented better.