OpenFF Theory Benchmarking Sets do not use correct OpenFF "default" QCSpecification

[trevorgokey]

The following datasets use the "default" name for its only QCSpecification, but this name is, for now, strictly reserved for b3lyp-d3bj/dzvp:

TorsionDriveDataset OpenFF Theory Benchmarking Set B3LYP-D3BJ def2-TZVP v1.0
TorsionDriveDataset OpenFF Theory Benchmarking Set B3LYP-D3BJ def2-TZVPD v1.0
TorsionDriveDataset OpenFF Theory Benchmarking Set B3LYP-D3BJ def2-TZVPP v1.0

These datasets should be resubmitted with the standard spec (v1.1). If each set has the same molecule input, then they should be collapsed into a single dataset. If it is desirable to keep the datasets separate, then all should be resubmitted with the proper default spec, and their associated auxilliary model.

Also, the original input input_json had a description field in each molecule. Did @hyejang want this information in the data? It appears it did not transfer during submission. Finally, likely of less importance, the entry indices were changed from torsion driven smiles to stringified integers.

[dotsdl]

Ah, this is something I didn't catch as the reviewer. Apologies all. I will take this one on and create a new dataset that subsumes these multiple specs into a single submission for @hyejang.