Fix bow model

[Irozuku]

This pull request updates the tokenization logic in the bag-of-words text classification model to ensure that all input columns are removed after tokenization, rather than just the "text" column. This change improves compatibility with datasets that may have different or multiple column names.

Improvements to dataset preprocessing:

• Updated the fit method in bow_text_classification_model.py to remove all input columns after tokenization by passing x.column_names to the remove_columns argument.
• Updated the predict method in bow_text_classification_model.py to similarly remove all input columns after tokenization.