eigenmode phase convention: choose real transverse fields for zero-thickness case

[stevengj]

Currently, our eigenmode solver chooses the phase to maximize the norm of Re(H), corresponding to fix-hfield-phase in MPB.

However, for a zero-thickness (i.e., constant-cross-section) eigenmode solve, there is a more common convention: choose the phase so that the transverse fields are purely real. It should be fairly easy to change our phase convention to match this case.

See also #2409