This PR resolves #491, resolves #514, and makes a lot of progress on #495.

This PR adds:

1. error-weighted optimal photometric extraction using a somewhat similar method to our spectroscopy code. I currently do not have code to mask pixels that deviate too far from the spatial profile; let me know if you think that's critical to add at this stage.

2. aperture photometric extraction using the open-source photutils, which allows for different aperture shapes, different aperture edges (allowing for "center" edges where a pixel is either inside the aperture or outside the aperture depending on whether the center of the pixel lies within the aperture, or "exact" edges where each pixel is weighted by the fractional area of the pixel that lies within the aperture), and is a fast and performant package.

This PR also greatly speeds up:

1. the flagging of bad pixels in photometry data by re-using the nircam.flag_ff function

2. the interpolation-based replacement of bad pixels for photometric data by adding multiprocessing support

I also restructured the photometry code to run each step for all integrations before moving on to the next step. This makes parallelizing easier and makes some steps easier.

I also fixed a bug with meanerr which was causing np.nan values to be returned for skyerr. There were also a couple other bug patches throughout.

Things remaining to do before the PR is ready for review:

• Tidy up the apphot.apphot function inputs and outputs to be more like apphot.photutils_apphot
• Add support for aperture_edge='exact' when phot_method='optimal'
• Manually test phot_method='poet' and check that it is as good or better than before
• Manually test phot_method='photutils' and check that it performs at least similar to poet
• Manually test phot_method='optimal' and check that it performs at least similar to poet
• Add documentation for new S3 meta parameters
• Make sure all new functions have proper docstrings
• Fix figs 3306 and 3504 for new aperture shapes