Add pandas workaround back in

[davesque]

Pull Request Description

We realized earlier today that the pandas workarounds that were removed by #1677 and #1679 are probably still needed as the bug reported in pandas-dev/pandas#22501 may still exist. More investigation is needed to understand exactly what causes data frames with misaligned categorical indices to be merged incorrectly. For now, we should at least revert the changes made by the two aforementioned PRs.

Left to do:

• More investigation into what causes the categorical merge bug to show up
• Regression tests that check for improper merges in _calculate_agg_features resulting from the categorical merge bug
• Add more details in this ticket about the code that revealed the issue

[rwedge]

@davesque this can merge into main, woodwork-integration branch has already been merged

[codecov]

Codecov Report

Merging #1731 (9e72515) into main (1909f2d) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##             main    #1731   +/-   ##
=======================================
  Coverage   98.69%   98.69%           
=======================================
  Files         138      138           
  Lines       15361    15368    +7     
=======================================
+ Hits        15161    15168    +7     
  Misses        200      200           

Impacted Files	Coverage Δ
...s/computational_backends/feature_set_calculator.py	98.44% <100.00%> (+0.01%)	⬆️
featuretools/entityset/entityset.py	99.21% <100.00%> (+<0.01%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1909f2d...9e72515. Read the comment docs.

[thehomebrewnerd]

@davesque No issues with the reverted changes, but just a couple comments:

1. Have you been able to confirm that reverting these changes fixes the bug you were seeing?
2. Can we create an issue to investigate this further and try to add tests to our test suite that will fail if remove this code with the current pandas version?

[davesque]

@thehomebrewnerd

1. Have you been able to confirm that reverting these changes fixes the bug you were seeing?

Yep, it fixes the issue. Confirmed that before making the PR.

2. Can we create an issue to investigate this further and try to add tests to our test suite that will fail if remove this code with the current pandas version?

There's a bullet point in the above PR description relating to this. I think the tests should be part of this PR.