Change definition of FluxTime and DFT fields with complex fields

[caseyflex]

This PR:

• Changes FluxTime and DFT calculation to only use real part of fields
• Updates sim.complex_fields based on nonlinear models

[momchil-flex]

Could you add a changelog item explaining the change?

I guess let's merge this once we're 100% clear on the modifications we want to do, including freq domain.

[momchil-flex]

So in the current state, the following would be happening:

• FieldTimeMonitor returns complex fields just in case, but we make it clear that the real fields would typically be the ones of interest, and the imaginary fields are not used in many derived quantities.
• FieldMonitor returns fields that are the DFT of Real(E(t))
• FluxTimeMonitor returns flux based on Real(E(t)) only; FieldTimeMonitor.flux is also correspondingly modified.
• Time and frequency domain fluxes are then consistent with one another and the total power is independent of whether real or complex fields are used.