Ewald sum simplified into modular functions; GCMC now works for rotatable molecules

[SimonEnsemble]

• Ewald sum was split to be more modular by computing short-range, long-range, self, and intramolecular interactions in separate functions. We still have maintained the performance of the last branch (see Ewald_test.jl for timing) despite needing to construct PointCharges from MOF atoms.
• NIST tests for electrostatics all pass!
• few extra tests to ensure rotations and translations are preserving bond lengths and adjusting the center of mass of molecules properly
• rotation move added to GCMC that is performed during translations to properly explore phase space
• nicer printing in GCMC function
• a molecule now interacts with its own periodic image per the new Electrostatics.jl function.

:heavy_check_mark: runtests.jl pass :heavy_check_mark: timing in Ewald_test.jl 6 ms for electrostatic_potential_energy call :heavy_check_mark: SBMOF-1 Xe adsorption GCMC tests pass :white_check_mark: CO2 adsorption tests in ZIF-71 pass at low pressure (still waiting for higher pressures).

[SimonEnsemble]

• neato; can preallocate with atoms = Array{Symbol, 1}(n_atoms)
• we can write a nearest_image!(dx::Array{Float64}, repfactors::Tuple{Int, Int, Int}) if a need arises later, but I think it will be faster if we don't have to pass in (1,1,1) every time we call it.

I think even

function nearest_image!(dx::Array{Float64}, repfactors::Tuple{Int, Int, Int})
end

nearest_image!(dx::Array{Float64}) = nearest_image!(dx::Array{Float64}, (1,1,1))

will incur a performance penalty? Since we call it literally millions of times I figured we should take advantage of the simplification?