Add "informative" set (subset) from Ehrman

[davidlmobley]

Jordan Ehrman's analysis of eMolecules pulled out molecules with minimized geometries which are substantially different in different force fields. His work is still finishing a final pass, but I can go ahead and provide a first batch (from the first portion of eMolecules). At present I have 9k compounds. After filtering for < 3 rotatable bonds it's 1,117 compounds with pretty diverse chemistry. I'll put in a PR for this set.

It would be good to run as an optimization dataset without fragmentation for our testing/benchmarking -- though lower priority than the Genentech and Pfizer sets. I'd also like to get one set of drug fragments into the queue ahead of it potentially.