Near2Far refactor phase 1: changes to storage and access of far fields

[shashwat-sh]

• Near2far data structures now store E,H rather than N,L
• Significant refactoring and simplifications to how fields are accessed and re-projected to new distances

Previous behaviour:

• All data structures had a fields method which returned fields in a coordinate system that depended on the monitor, e.g. spherical for the angle monitor, Cartesian for the Cartesian monitor
• Radar cross section method was available for only the angle monitor
• All these methods took an argument r to project fields to that given distance

New behaviour

• All data structures have the following methods:
  • fields_spherical
  • fields_cartesian
  • radar_cross_section
  • power
    so that all quantities are available for all monitors, and in both coordinate systems.
• By default, fields are projected to a distance of 1m.
• Re-projection to a different distance is no longer done with an argument. Instead, one explicitly calls renormalize_fields(distance) so that all fields stored in the data become renormalized to the new distance.
  • The field data structures' coordinates will be updated to remember this new distance. This way, one can re-project as many times as they want. Each time, the old phase is removed, the new phase is applied, and the field data coordinates are updated.

Rationale:

• Make things simple, explicit, and uniform for all three types of monitors.
• Unify the interface as much as possible in anticipation of exact projections, in which case everything can stay the same, except renormalize_fields won't be available because the phase is "baked into" the exact Green's function.

Backend PR: #557

[momchil-flex]

I think this makes a lot of sense now!