Tweak edges using a gradient method

[rcooke-ast]

This PR adds additional functionality to the tweak-edges algorithm. The current implementation is prone to error when there's vignetting (see below image for an example).

The top panel is the slit profile (black), and some vertical lines that show the initial, tweaked (threshold - already implemented in PypeIt), and tweaked (gradient - new method in this PR) determinations of the slit edges. The tweaked (gradient) is a new method that I’ve implementing that is based on the strongest gradients of the slit profile (see bottom panel)… the smoothed profile is in purple, which is what the tweaked locations is based on. The key problem with the old approach is vignetting.

I will soon be submitting another PR that refactors the spatial flexure to be (optionally) slit-dependent. This is the start of that effort, and I won't develop further on this until PR #1755 clears.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 27.33333% with 109 lines in your changes missing coverage. Please review.

Project coverage is 38.19%. Comparing base (14d810f) to head (510d2b1).

Files	Patch %	Lines
pypeit/core/flat.py	9.09%	40 Missing ⚠️
pypeit/flatfield.py	29.62%	19 Missing ⚠️
pypeit/coadd3d.py	0.00%	10 Missing ⚠️
pypeit/core/procimg.py	0.00%	9 Missing ⚠️
pypeit/core/datacube.py	0.00%	8 Missing ⚠️
pypeit/spectrographs/gtc_osiris.py	0.00%	6 Missing ⚠️
pypeit/images/rawimage.py	0.00%	4 Missing ⚠️
pypeit/spectrographs/gemini_gnirs.py	0.00%	4 Missing ⚠️
pypeit/find_objects.py	0.00%	3 Missing ⚠️
pypeit/alignframe.py	0.00%	2 Missing ⚠️
... and 3 more

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@                  Coverage Diff                  @@
##           kcwi_cube_updates    #1796      +/-   ##
=====================================================
- Coverage              38.22%   38.19%   -0.03%     
=====================================================
  Files                    208      208              
  Lines                  48256    48311      +55     
=====================================================
+ Hits                   18444    18451       +7     
- Misses                 29812    29860      +48     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[kbwestfall]

Looks good to me, so I'm approving now. As far as I can tell, you mostly remove the initial=True flags for the code relevant to the SlicerIFU only. Is that true? There was at least one case where I wanted to make sure that we understand changes to the default behavior.

[rcooke-ast]

Thanks for the review, @kbwestfall - all addressed!

[debora-pe]

This looks good for me in general. I'm approving.
But I have a question. The fact that you removed initial=True to all the code relevant to the SlicerIFU, does it mean that the parameter tweak_slits cannot be False for SlicerIFU? If it's correct, is that clear to the users?
Similarly, spatial_fit_finecorr() also have initial=False. Does it mean that it is always run only when tweak_slits=True? Is that clear to the users?

[rcooke-ast]

Thanks for the feedback, @debora-pe!

The fact that you removed initial=True to all the code relevant to the SlicerIFU, does it mean that the parameter tweak_slits cannot be False for SlicerIFU?

Just to confirm, tweak_slits can be False for SlicerIFU. These latest changes basically just make the code identical for MultiSlit and SlicerIFU when it comes to using either the initial or tweaked edges. The main difference is that MultiSlit and SlicerIFU can now use an alternative method to define the tweaked edges. Because the way slits.select_edges() works, the default will be that the tweaked edges are used, and if those are not available (or if the initial edges are specifically requested) then the initial edges are used.

Similarly, spatial_fit_finecorr() also have initial=False. Does it mean that it is always run only when tweak_slits=True?

Like the above response, the spatial_fit_finecorr() function uses onslit_tweak, but this variable is defined by using the tweaked edges by default, and if those are not available, then the initial ones are used.

So, the full functionality is still available to all users - thanks for checking 👍

[debora-pe]

Thank you @rcooke-ast for the explanation!