Colab Notebook development

[sgpjesus]

Changelog

These changes accommodate several QoL improvements to interact with the notebooks and provide a better user experience.

These are:

• Creation of the class DefaultExperiment, an experiment that allows the users to pass their dataset configuration, as well as the size and group of methods they want to test (aequitas/fairflow/experiment/default.py; this includes default configurations in aequitas/fairflow/experiment/_configs.py).
• Separation of BaseEstimator methods from InProcessing methods. These estimators include methods that are traditionally fairness-blind (e.g., Decision Trees, Neural Networks, k-nearest neighbors, etc.). BaseEstimator methods include a parameter unawareness not to use the sensitive attribute as a feature for prediction.
• Restructure of the aequitas.fairflow.methods directory.
• Creation of the class LabeledFrame (introduced in the Datasets), an extension of pandas.DataFrame with the attributes X, y, and s, for easier access to these specific columns.
• Change GenericDataset to support splitting, either randomly or depending on a categorical variable.
• Update both plots to have an easy-to-use wrapper.
• Add a method to utils for reading and restructuring the results of an experiment.
• Add a method to utils for downloading the example files from the repository, used in colab notebooks.
• Several small updates, such as updating imports with isort and formating the code with black.