OPEN-3410 Support predictions upload

[gustavocidornelas]

Adds support for the predictions upload via the Python API.

Summary

• Predictions are uploaded together with a dataset.
• On the add_dataset and add_dataframe methods, the user can specify a predictions_column_name as an arg.
• On the actual dataset, the column with the predictions should contain lists with the class probabilities. For instance, for the Churn dataset, it should look like:
  
• The validations done for predictions are:
  1. Check if predictions_column_name is a column in the dataset;
  2. Check if all values in predictions_column_name are lists;
  3. Check if all lists in predictions_column_name are of the same length;
  4. Check if the length of the lists in the predictions_column_name match exactly the number of classes specified in class_names;
  5. Check if the sum of the values of the individual lists with class probabilities is equal to 1 (I'm allowing an error margin of 10%, so I allow if it's between 0.9 and 1.1).
• All the 27 states of commit bundle were outlined on this Notion doc. A warning is thrown on the cases where the bundle could be ambiguous.

[linear]

OPEN-3410 Support predictions upload