Continuous integration t̶e̶s̶t̶i̶n̶g̶ linting

[pjf]

We already have some tools which are able to do code linting, spotting references to non-existent reactions, or single-use reaction classes, or potentially other problems.

It'd be great to:

• Have a codebase which lints free of major errors.
• Use travis to check to see if commits or pull requests introduce obvious errors.

The first part is the hard part. I've got lots of travis experience, and deploying it would be easy.

Alternatively, we can do something magical that records which linting errors we've seen, and insists we don't introduce new ones. But that's adding a lot of complexity, and I'd really just love to fix the errors that we already have (or mark some error types as "soft" errors that we're willing to let past).