Including different paradigms and accounting for variations

[Div12345]

Any steps necessary to including the ability to handle other paradigms including extensions to different forms of ERPs? Should any measures be done to also indicate variations in the experimental conditions in which the dataset has been obtained? I see that #87 mentions this partly and for being able to do #93 to include own datasets, this should be an important part as well maybe?

[jsosulski]

As @vinay-jayaram suggested in #65, a more flexible approach to dataset attributes would be needed to handle e.g. conditions nicely. For my use cases, just adding a conditions attribute to the base dataset class would be sufficient. This could be None or ["default"] for all currently implemented datasets meaning that there were no different conditions.

However, additionally we need to handle conditions in the paradigm definitions where we get data. For instance, in paradigms/base.py the get_data function stores only subject, session and run information in a data points meta data. When condition information is also present, the evaluation classes could handle conditions nicely, e.g. "pool across conditions" "evaluate within conditions".