Refactor how Featuretools 1.0.0 compatibility was handled

In #2848 , we went forward with an intermediate solution to maintain compatibility with the new featuretools v1.0.0 and nlp-primitives 2.0.0. This involved a deletion of a woodwork accessor in the DFSTransformer within the EvalML code. We want to revisit this after the related [Featuretools issue](https://github.com/alteryx/featuretools/issues/1740) is addressed.

One possible solution is to request that the Featuretools team handle the addition of dataframes to an entityset that possess no name but already have a valid woodwork schema (have been initialized). Another possible, but hackier, solution is to check for woodwork initialized dataframes without a name and give them a name.