properly resolve interfaces, clean cruft

* Circular nuclei dissolve for halfwidth < 3.875*dx
* Zero-flux means the gradient on the wrong side of the boundary
  is zero, not that the divergence should be off-centered
* Store parts of the Eqn. of Motion in the grid for post-mortem
* Wrap GSL fsolve and interp2d in classes for readability and resusability
* Separate GSL class definitions from implementations
* Template member functions, not classes
* Drop references to parabolic free energy, which was definitely wrong
* Be VERY careful calling an object inside an OpenMP loop!

Author: Trevor Keller