Improve routines for scaling numeric/categorical column values (across Pandas and Polars dfs)

[rich-iannone]

The current implementation of scaling numeric and categorical column values (needed for generating gradients and obtaining specific color values) could be more robust to less common value types in Pandas and Polars DataFrames. The relevant code is in here: https://github.com/posit-dev/great-tables/blob/287ba5202a6ab97d74d8aad65cf92d4d4c17dc48/great_tables/_data_color/base.py#L529.