put data required for testing into separate package about mdanalysis

[GoogleCodeExporter]

Ideally we would something like the MDAnalysis-X.Y.Z package that contains the 
code (including the test cases under MDAnalysis/tests because they need to be 
in sync with the rest of the code) and MDAnalysis-testdata-X.Y.Z.

Come considerations below --- comments are very much appreciated!!

- I think that the testdata package should have the same release number as the 
main package that it was last updated for. As long as there is not a newer 
testdata package one uses the one with the highest release number and it should 
work.

In this hypothetical example below, testdata-0.7.4 does not change until 0.8.0 
and is used throughout the remainder of the 0.7.x releases:

MDAnalysis 0.7.4   0.7.5   0.7.6   0.7.7   0.8.0
testdata   0.7.4-------------------0.7.4   0.8.0

- Ideally I would have the testdata package install the trajectories etc in the 
same place as where we have them right now (MDAnalysis/tests/data) although we 
could probably live with another place by changing paths in tests/datafiles.py 
or making datafiles part of the testdata package.

- Python name for the package? MDAnalysisData? 

- Is there a way to build two python packages from a single source tree (using 
setuptools or another system) or do I have to make a separate SVN branch and we 
essentially maintain two packages?

Original comment by orbeckst on 29 Jul 2011 at 5:55

[GoogleCodeExporter]

Split off a branch MDAnalysisTestData at r860. It builds a separate package 
(named MDAnalysisTestData ...) that contains the trajectories required for 
UnitTests. From 0.7.4 onwards it MUST be installed separately to MDAnalysis in 
order to run the tests.

At r866 the trajectories were removed from the trunk. MDAnalysis 0.7.4 will be 
a much leaner installation package than before.

Original comment by orbeckst on 7 Sep 2011 at 8:26

• Changed state: Fixed