By default the jenkinsapi Jenkins class will trigger an eager loading of
all Jenkins jobs via the API. Our Jenkins server buckles under the
weight of this and so we do not want to do this until we need to and
ideally future changes will replace unfiltered polling completely.

There's a malinteraction with the jenkinsapi library and Python
convention which has been causing extremely large HTTP requests whenever
the truthiness of an instance of the Jenkins class is tested.

Jenkins defines a `__len__` method which is used to determine
truthiness. But this method triggers a fetch of every job on our Jenkins
host which can take 2 - 5 minutes to return due to the sheer volume of
jobs.

By avoiding this in favor of an explicit check whether jenkins is an
instance of the JenkinsProxy class we can save taking a really big
runtime hit and cut down on spurious failures due to timeouts.