Finish testing integrators and swap default integrator

[davidlmobley]

In #50 and (to a lesser extent #53 ) we've been discussing changing the default integrator to NonequilibriumLangevinIntegrator. We need to do so, but first we need to finish testing:

@sgill2 has some questions he needs to answer (and re-test if needed) which are asked here: #50 (comment) (and could be answered on this issue) for tests on lysozyme, I believe.

@nathanmlim is going to test on a soluble epoxide hydrolase (?) system, cross-comparing this with the current integrator and report back.

Once we've tested a bit more carefully and documented the results it sounds like we should be able to switch, which will resolve issues with code duplication discussed in #50 and will hopefully also improve acceptance, etc.

[sgill2]

Closed via #89. Testing showed that the new AlchemicalExternalLangevinIntegrator performed better than the old VV integrator.