"missing" roi_id column when applying classifier to new data according to the example notebook

[MaksHess]

Kelvin ran into an issue where he could not apply the classifier to new data according to this notebook: simple_classifier_example.ipynb

The quick fix was to add a roi_id column to the dataframe before calling predict.

...
# Load classifier and dataframe
clf.predict(df)
# fails
clf.predict(df.assign(roi_id='string_column_that_is_unique_when_combined_with_label'))
# passes

Since we don't really care about the index for inference we should ease the schema requirements.