Segfault workaround

[jchodera]

Attempting to work around openmm segfaults.

[rmcgibbo]

Is there an issue/fix related to this upstream in OpenMM?

[jchodera]

Yes. We're trying to create an isolated test case.

Here's an example of the behavior we're seeing in yank:
https://travis-ci.org/choderalab/yank#L1268

Testing reference/alchemical overlap for Lennard-Jones fluid without dispersion correction... ... /home/travis/miniconda/conda-bld/test-tmp_dir/run_test.sh: line 3:  3650 Segmentation fault      nosetests yank --nocapture --verbosity=2 --with-doctest

[rmcgibbo]

Yes. We're trying to create an isolated test case.

Got it.

[jchodera]

Created OpenMM issue openmm/openmm#781 with minimal test case.

The fix I've put in place reduces the default density for LennardJonesFluid and LennardJonesGrid to be appropriate for a gas or dilute fluid. This should avoid the segfaults in yank's travis.

Should be good to merge once tests pass.

[jchodera]

I've removed all the openmmtools packages from the binstar and restarted the Jenkins build, but we should probably just increment the minor version number in the future. Sorry for forgetting to do that!