Improve performance of feature_columns_numeric and feature_columns_categorical

[cachafla]

Internal Notes for Reviewers

External Release Notes

[johnwalz97]

love it!

[github-actions]

PR Summary

This pull request optimizes the method for detecting numeric and categorical feature columns in the validmind/vm_models/dataset/dataset.py file. Previously, the code loaded data into memory to determine the data types of feature columns. The updated implementation retrieves data types without loading data into memory, improving performance and efficiency.

The changes involve:

• Using dtypes to get the data types of feature columns directly.
• Applying pd.api.types.is_numeric_dtype and pd.api.types.is_categorical_dtype to determine numeric and categorical columns, respectively.
• This change reduces memory usage and potentially speeds up the process of setting feature columns.

Test Suggestions

• Test the _set_feature_columns method to ensure it correctly identifies numeric and categorical columns with a variety of datasets.
• Verify that the method performs efficiently with large datasets, ensuring no significant memory usage increase.
• Check edge cases where feature columns might have mixed data types or unexpected data types.
• Ensure that the method behaves correctly when there are no numeric or categorical columns present.