Improve trajectory reading ability

Closes #51 (closed)

The implementation of the new trajectory reader and the accompanying tests now seem to work. Other related changes:

• Changed how trajectory reader is chosen based on new required keyword trajectory_format in the API.
• Added optional keywords lengthunit and timeunit for when these cannot be read from trajectory, or should be overwritten for some reason. Based on these, proper conversion to the internal dynasor units (nm, fs) can be done when reading the trajectory
• Removed old (and, to the best of my knowledge, non-functioning) trajectory readers based on libgmx and molfile plugin.

To do before merging:

• Test performance and compare to the other readers. Do you want to do this @erikfransson? See #52
• Make pipeline run by making MDAnalysis available. How do I do this? Locally, where I have pip installed MDAnalysis, the tests run without any issues.
• Should we cite MDAnalysis somewhere?