get_farfields routine to obtain far fields as NumPy array

[oskooi]

Currently, there are only two routines available for obtaining the far fields:

1. get_farfield: returns the far field at a single point
2. output_farfields: writes the far fields in a given Volume to an HDF5 file

It would be useful to have an additional third routine which returns the far fields in a given Volume as a NumPy array. This would facilitate post processing directly from within a single Python script without having to write and read from disk.

As an extension of this feature, rather than restrict the far field output to a rectilinear Volume, the region could be generalized to a GeometricObject (e.g., Sphere, Prism, Cylinder, etc.)

[stevengj]

Looks straightforward, because the dft_near2far::save_farfields function already collects all the necessary information in an array before writing it to disk. So we would want to:

1. Refactor save_farfields into two routines — a get_farfields_array that collects the EH array and a save_farfields that calls get_farfields_array and then saves it to HDF5.

2. Change get_farfields_array to use sum_to_all instead of sum_to_master so that all processes get the data.

3. Make a high-level wrapper around get_farfields_array and call it from Python.