Prove convergence of equilibrium solve

[bclyons12]

We should determine the order of convergence in the number of radial finite element, poloidal Fourier modes, and MXH coefficients. This would certainly be needed for any peer-reviewed publication. It would like be best to do a Solovev case so there is an analytic answer to compare to. Some initial thoughts from cursory testing.

• I'm seeing convergence in ρ finite elements, though the order of convergence should be tested, as should the scaling to very high finite element number.
• There appears to be some coupling between the number of θ Fourier modes (M) and the number of MXH modes (L). If M < L, the code can struggle to converge, while if M > L, I don't see significant improvement in the solution. This kind of makes sense. You can only ever eliminate L oscillations in your solution by refitting the MXH coefficients, so the flux surfaces won't converge away errors high Fourier mode errors for M > L. It's possible we'll just want to fix M = L.

To-do:

• Measure order of convergence. I think the error should go like h^4.
• Compute matrix of errors for M and L. I expect for fixed L you'll see that for M < L you get high error and for M > L you'll see little improvement. Going down the diagonal, you'll get a meaningful convergence.