Add pandas 2.0 support

[thehomebrewnerd]

Add pandas 2.0 support

Closes #2532
Closes #2556

[codecov]

Codecov Report

Merging #2585 (dccd005) into main (2687140) will increase coverage by 11.73%.
The diff coverage is 100.00%.

@@             Coverage Diff             @@
##             main    #2585       +/-   ##
===========================================
+ Coverage   87.74%   99.47%   +11.73%     
===========================================
  Files         393      403       +10     
  Lines       23271    24198      +927     
===========================================
+ Hits        20419    24071     +3652     
+ Misses       2852      127     -2725     

Impacted Files	Coverage Δ
...ndard/transform/binary/multiply_numeric_boolean.py	100.00% <ø> (ø)
...ive_tests/transform_primitive_tests/test_season.py	100.00% <ø> (ø)
featuretools/demo/flight.py	96.10% <100.00%> (+75.32%)	⬆️
featuretools/synthesis/encode_features.py	98.79% <100.00%> (+14.64%)	⬆️
.../tests/computational_backend/test_dask_features.py	100.00% <100.00%> (ø)
featuretools/tests/entityset_tests/test_es.py	100.00% <100.00%> (ø)
...ools/tests/entityset_tests/test_last_time_index.py	100.00% <100.00%> (ø)
...aggregation_primitive_tests/test_agg_primitives.py	100.00% <100.00%> (ø)
...ggregation_primitive_tests/test_num_consecutive.py	100.00% <100.00%> (ø)
...s/tests/primitive_tests/test_transform_features.py	99.87% <100.00%> (ø)
... and 2 more

... and 52 files with indirect coverage changes

[sbadithe]

This is not a blocker (more of an incidental comment), but seeing the if clause here makes me consider if we should eventually refactor this test. Given that the previous data objects are being cast to float64 dtype as well, and the test previously did not require if clauses, it seems like there are three separate classes of behaviors here:

• How the kurtosis method handles the int64 dtype
• How the kurtosis primitive handles the float64 dtype when the data "looks like" an int64 (evidently this doesn't seem to differ from the previous class)
• How the kurtosis primitive handles the float64 dtype when the data are "true" floats

Separating those at some point might help to better document this primitive's behavior.