This allows now to compute (the average of) a x-dependant bias.

Based now on JSON files in results folder. evaluate_tabular.py has been
renamed into evaluate_metrics. JSON files have also been updated. Need
to check whether correct now for all datasets.

Corrected eiv_prediction_number_of_draws in JSON configuration files
and updated the intervals of std_y updating for some of the datasets.

Previously, this was done by looking at the average of the predictions
over the posterior predictive. This was removed, as this didn't make
much sense.

The evaluate_tabular script was simplified to ease the analysis of more
quantities. Several coverage quantities and the bias were added to these
quantities. However, they perform rather poor.

Added EiV training scripts for the three datasets and moreover included
bias evaluation in `evaluate_tabular`. The inclusion of some coverage
measure is still needed.

This covers all regression datasets treated in the MC Dropout and Deep
Ensemble paper. Results are comparable or even better. For multivariate
dataset, the decouple_dimensions keyword in the evaluation scripts can
be used to follow the (rather weird) convention of these papers.

Added training and testing for nonEiV-model on the power plant dataset.
The evaluation script for MSD was also added.

For NLL metric, weird convention from MC Dropout was integrated.
Power Plant example missing.