Create a CI job that runs Blinkr against the production site.

• Ensure the job runs in the US. The current Blinkr job runs in Singapore and suffers from too many false positives caused by the high latency connection to the US. A private OS1 Jenkins slave could be a good place to run this.
• Initially run the job every 24hrs.
• Update Blinkr to produce a JUnit report that can be processed by Jenkins.
• Somehow configure the job to only fail on 'Danger' issues. The danger list should be down to zero. Getting the other items (e.g. searchisko warnings) down to zero will be a long-term task. We don't want the job always failing, otherwise it will not get used as much.
• Create a Docker image to run Blinkr and make it available to run from control.rb. By default it should test the local build of the site; but we would also need a configuration to test other builds, such as the production build.