Update to work with dask 2022.4.2 and pyarrow 8.0

[zaneselvans]

• Set PUDL_INTAKE_CACHE to the default value if it is found unset.
• Use index=False and split_row_groups=True in the tests and examples when reading from the catalog, since the monolithic file is huge and otherwise it'll end up in a single dataframe.
• When comparing actual and expected dataframes in the integration tests, don't require CategoricalDtypes to be identical, since the known categories include only ID in one dataframe and all the states in the other (since they're part of the parquet metdata?) even though the data is identical.

[codecov[bot]]

Codecov Report

Merging #19 (426e737) into main (5869403) will not change coverage. The diff coverage is 100.0%.

@@           Coverage Diff           @@
##             main      #19   +/-   ##
=======================================
  Coverage   100.0%   100.0%           
=======================================
  Files           2        2           
  Lines          41       42    +1     
=======================================
+ Hits           41       42    +1     

Impacted Files	Coverage Δ
src/pudl_catalog/__init__.py	100.0% <100.0%> (ø)

---

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 5869403...426e737. Read the comment docs.