Residuals fit improvement in photometry script

[PK0207]

Describe the bug
Photometry work by fitting a function to a Delta M - M plot, where Delta M is the difference between the instrumental and gaia magnitude for the reference stars, and M is the instrumental magnitude of the reference stars. As of right now, most of the fits are not great, but there is one that is.

We want the fits to consistently have the quality of the better fit, so we need to figure out a way to do so.

To Reproduce
Steps to reproduce the behavior:

1. Go to 'The Lightcurve Project Drive'
2. Run the notebook called photometry.ipynb (runs in jupyter and vscode)
3. Scroll down to cell 8 #Convert gaia g, r magnitudes to...
4. See resultant residual plots

Expected behavior
We expect the fits to be of the quality of better fit shown above, and robust to outliers in general. If all the fits are improved, we expect the errors on our zero point (and therefore the quality of our zero point calculation) to improve.

Ideas
There are currently two fixed parameters in the residual fit, gain, and fwhm of the stars. The gain is fixed for an image, and can be pulled from the header. The FWHM of the star's PSF is also fixed, assumed to be constant for all the stars but this might not be the case. We might improve this in some of the following ways:

1. Take the mean of the fwhm of the reference stars. I'm not sure if this parameter is being claculated by sep or photutils anywhere.
2. Currently we conduct the fit by minimizing the median absolute deviation of the function. This fitting algorithm could be changed, but this is not the first approach to take.

Desktop (please complete the following information):

• When was photometry.py last run? 01/22/23
• What data is being used? LC Project drive linked at the beginning of this issue

Additional context
This is before the integration of photometry into the pipeline. This is part of the effort to do so.