Multi-Score Paradigms

[davisethan]

Problem: MOABB assumes paradigms use one score for evaluations. However, besides discrimination (e.g. accuracy, AUROC) measuring calibration (e.g. NLL, Brier score, ECE) is also useful.

Proposed Solution: Allows custom paradigms to use multiples scorers. The parameter additional_columns from evaluations is used for multiple scores in results. The column score in results is kept and used as a register for statistical analysis and plotting.

Google Colab: https://colab.research.google.com/drive/1gFDBIfdbAWl2UdvslxVnLFhV7qws6_qm?usp=sharing

[davisethan]

#948