Add timeouts for stubs

Allow runners to time out tests if a stub takes too long to finish. The functionality could be built into the runners so all stubs don't have to start implementing timeouts one by one.

If implemented should the result in those cases be `ERROR`? Or should we reserve `ERROR` only for actual stub errors and add a new `TIMEOUT` type (or somesuch)?
