Replace pandas IO in margin generation

[camposandro]

In the margin generation pipeline, read and write parquet files using pyarrow.parquet instead of pandas's read_parquet.

This allows us to manipulate the margin schema directly with pyarrow tables, and it solves the schema issues we were facing when generating margins for catalogs with nested columns. I added a unit test where I created small_sky_nested_catalog, from small_sky_object and small_sky_source, which covers this use case. Closes #465.

• My PR includes a link to the issue that I am addressing

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 94.82%. Comparing base (cd17800) to head (673a41f).
Report is 11 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #500      +/-   ##
==========================================
- Coverage   94.97%   94.82%   -0.15%     
==========================================
  Files          28       28              
  Lines        1811     1818       +7     
==========================================
+ Hits         1720     1724       +4     
- Misses         91       94       +3     

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

[delucchi-cmu]

Looks great!