Make simple but nontrivial test system with fast run time

[davidlmobley]

We've needed a simple test system for basic BLUES which will allow us to (with minimal simulation effort) verify that rotational moves are in fact working correctly and yielding correct answers. We probably want a system we can in principle run to convergence with standard MD (to get the right answers which would be archived for comparison) but which will run very fast with BLUES.

In discussions we came up with an eight-atom test system we think may work, which would consist of a "ligand" made up of a two atoms connected by bonds (with very weak partial charges -- a slight dipole) in a "binding site" made up of six atoms at the centers of the faces of a cuboid (like a rectangle but 3D), with the end atoms slightly charged (so that the ligand has a preferential orientation). The cuboid would be somewhat longer than it is tall and wide so that the "ligand" primarily fits end-wise, with one direction being preferred over the other somewhat.

The dimensions of the box would need to be tuned carefully so that transitions between "binding modes" are possible but slow with standard MD, whereas with rotational moves it will be easy to switch between them. The purpose of the charges is just to break the symmetry slightly so that of the two "preferred" orientations, one has substantially higher population than the other.

@sgill2 was going to set this up.