Add support that dataset entry points have resource name "learningData"

Once these are merged:

• https://gitlab.datadrivendiscovery.org/MIT-LL/d3m_data_supply/merge_requests/14
• https://gitlab.datadrivendiscovery.org/d3m/datasets/merge_requests/16

We should make sure when we are reading datasets we make those resources with learningData as name dataset entry points. Only if no resource has such name, we check resource filename if it is "learningData" (for backwards compatibility).