Validation/processing of QCArchive records

[mattwthompson]

Description

Drafting part of the backend for automated benchmarking. The scope here is to look at a QCArchive record and make an educated estimate of whether or not the output from quantum chemistry is sufficiently different from the SMILES/CMILES representation to invalidate it for benchmarking purposes. There will be many ways of doing this; my first pass is just throwing the OpenEye toolkit at it and seeing if something sensible comes out.

Most of the discussions for context have happened elsewhere, but https://github.com/openforcefield/cmiles/issues/36 is the sort of case we are looking to catch (among others).

Todos

Notable points that this PR has either accomplished or will accomplish.

• [ ] Split out and capture other "failures" as a set of validation flags
• [ ] Allow user overrides of default values of validation flags
• [ ] Add test cases of the molecules that trigger the above
• [ ] Don't rely exclusively on OpenEye
• [ ] Construct a spreadsheet/database that ties records to their validation results as booleans, to prevent redundant calculations
• [ ] Use the above data to skip validation (?)

Questions

• [ ]

Status

• [ ] Ready to go

[mattwthompson]

CI for this idea won't pass until QCSubmit is on Anaconda or I hack together a temporary workaround. I will also be adding an example notebook that shows off this approach