As everyone else realises how awesome XUnit support is with Jenkins, we need to refactor the code into something more sensible.

My plans are:

• an XUnitRerportBuilder class which has convenient methods like .error(), .ok(), .skipped()
• Proper XML construction rather than worrying about escaping various snippets.

XUnit and logging messages are increasingly filling very similar roles, so maybe there's a way we could handle both in one place? E.g. how lint_ctx gets tossed around, maybe we need to have a replacement lint_ctx which talks to TWO backends, the XURB backend and the planemo.io backend, dispatching messages to their appropriate locations.

Would that make sense @jmchilton? Are there places where this wouldn't be appropriate? Just wanted to have a short engineering discussion in this issue :)