Use pyarrow-backed DataFrame with read_csv

[gsheni]

• Use PyArrow backed data types for at reading data since PyArrow can be significantly faster and more memory-efficient than NumPy
• Starting with pandas 3.0 will change the default for strings to PyArrow strings (such as when calling read_csv).
  • Source

[WillKoehrsen]

@kmax12 Is this still something we want to do? Should be relatively easy to make the changes.

[kmax12]

i think all else equal, worth doing. not sure if it'll have a huge impact on us right now, so if it causes downstream problems, we can punt for now