Fix column replacement logic in converter job

[Irozuku]

This pull request updates the logic for rebuilding datasets with transformed columns in the converter_job.py file. The main focus is on improving how replacement and new columns are determined when transforming scoped columns.

Column replacement and filtering logic:

• The mapping for column replacement now uses scope_column_names directly instead of sliced transformed columns, ensuring replacements are based on actual column names.
• The replacement mapping is further filtered to include only columns present in both the transformed and original columns, improving robustness.
• The identification of new columns is updated to select columns from transformed_cols that are not in scope_column_names, making the logic clearer and more accurate.