r_ij consistency

[Eric Harper]

Fix consistency of all r_ij calculations in freud:

r_ij = r_j - r_i

Also, do not create local variables unless necessary so:

delta = m_box.wrap(m_points[j] - m_points[i]);

is preferred

[Eric Harper]

I think this is better, still needs to be fully checked

[Eric Harper]

Resolving, as I believe I fixed most of it. Still fix on case by case basis.

[Eric Harper]

I believe this to be resolved.