Using a full line-list of arc-lamp lines, try and rebuild XShooter static calibration frames with more lines

[thespacedoctor]

• [x] clean up the create_dispersion solution code ... from bottom up
• [x] use the RMS predicted vs measured location deviation for the full MPH line-sets to remove poor MPH sets (prior to polynomial fitting)
• [x] plot line fluxes, and attempt clipping on flux
• [x] clip full MPH sets post metric clipping, or
• [x] convert the clipping code to clip on MPH positional scatter, peak flux, total flux, fwhm_px
• [ ] fix the QC plotting code for SPH frames
• [x] test polynomial clipping on full MPH lines. Allow clipping on single or sets.
• [ ] add in boosted atlas ... can I improve the dispersion solution?
• [ ] play with map_to_image_displacement_threshold with more accurate lines ... might be able to increase and speed the convergence to a solution
• [ ] write the final enhanced line-list to a FITS binary table and replace the previus static calibration line-list

[thespacedoctor]

I have curated a very accurate, full-line atlas for each Ar Hg Ne Xe pen-rays used by Xshooter and SOXS from NIST and other sources. I am using the list to regenerate a more precise static line list for Xshooter, but the primary driver is to have the code in hand to do the same for SOXS when we get the NIR MPH data.