DOC Add Exponentiated Gradient in reductions.rst

[Hrittik20]

Issue: #1263

[MiroDudik]

Instead of encapsulating inside get_expgrad_models_per_epsilon, let's just present this as a script. First, you could just call exp_grad_est.fit(). Then report (or possibly plot) the (test or train) performance (in terms of accuracy and fairness) of the resulting ensemble, based on the predictions produced by exp_grad_est.predict().

Then in the second part of the example, you can extract predictors exp_grad_est.predictors_ and plot fairness vs accuracy of the individual models in the ensemble using plot_model_comparison()--perhaps also adding the ensemble predictions.

[romanlutz]

Note: use rest syntax, like elsewhere in this file for classes.