Improve choice of the best preferred identifier for conflated chemicals

[gaurav]

We should come up with strategies for improving choosing the best preferred identifier for conflated chemicals (something better than the smallest identifier).

Some ideas:

• Shorted preferred name or shortest synonym if it’s a direct subset of something else
• If cliques have SMILES — choosing the clique with the smallest number of components or the smallest molecular weight would help with
• Information content? Degree — number of linkouts?
• RxNorm active ingredient information

[gaurav]

Duplicate with #341, but I'm not sure which one to close.