Weak lensing DC2 updates

[timwhite0]

• Updated load_image_and_catalog and generate_cached_data. More efficient splitting of cached .pt files, and also more similar to the parent DC2 class than it was before.
• New baseline shear estimator that uses conv2d to compute a weighted average of ellipticities from neighboring tiles. The previous approach, which was to estimate shear in each tile by taking the average of ellipticities in that tile, can be recovered by setting the size of the kernel in conv2d to 1. After a bit of exploration, it seems like a kernel size of 5 or 7 provides the best approximation of shear. This likely has implications for the "optimal" receptive field of our lensing encoder.
• Renamed and reorganized some notebooks and other files.
• Updated catalog so that it only includes ~3.4 million galaxies with r-band flux ≥ 200 (r-band mag ≤ ~26). Previously our catalog contained ~10-11 million galaxies with r-band flux ≥ 50. With the new catalog, average ellipticity is not quite as good an estimator of shear as it was before, and definitely not as good as it was when we were (erroneously) including all ~128 million CosmoDC2 galaxies. See the bottom of ellipticity.ipynb for an example.

[jeff-regier]

Looks good. Please try to remove the duplicated code before merging.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 81.25%. Comparing base (e536606) to head (bcf827f).
Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1067      +/-   ##
==========================================
+ Coverage   81.23%   81.25%   +0.01%     
==========================================
  Files          23       23              
  Lines        3049     3051       +2     
==========================================
+ Hits         2477     2479       +2     
  Misses        572      572              

Flag	Coverage Δ
unittests	81.25% <100.00%> (+0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

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