Ensure empty data in query response str fields work with pandas>=3

[alyssadai]

Before pandas 3, string columns with some none-type values would be typed as object, but with pandas 3+, they are typed as str, with missing values being converted to np.nan. This breaks the type checking for some our current Pydantic models for Python>=3.11 (where pandas 3+ is used by default).

I think a long-term fix would be to normalize missing values to None to ensure they are not interpreted as numeric, but for now I have pinned the pandas version.

Originally posted by @alyssadai in #598 (comment)