Changes:

1. Moves the EOF offset calculation into standalone functions. This can either be restricted to just instrument or include a wrapper call in Forward in the style of e.g. xa.
2. The EOF offset is only applied two places now in the principle OE loop. One is on the output to the calc_meas call. The second is in dmeas_dinstrument.
3. AOE now uses the fm.eof_offset function to calculate the offset at a given super-pixel value. The "shifted" measurement is calculated in inverse_simple instead of analytical_line.py
4. A small edit on the indexing within analytical_line.py, although this requires a dedicated PR and some work to better abstract and add in the ability to "smooth" any arbitrary state-vector element.

I did minimal testing. I used an example EOF file David provided to ensure that the EOF scale values moved around during inversion. I can't comment on the quality of the solutions. I also haven't deployed at the scene scale. I can test to confirm the AOE changes if need be.