Reworks most of the HAProxy reloading code to be a little more testable (and adds tests), while keeping the logic mostly identical. Also adds the features in #158.

The major affect of all this is that

• We no longer exit if there is a problem trying to reload the haproxy config
• We have better logging in the area
  • The names of the statsd metrics were changed (well, I added a 2 and changed 1)
• We have a nice reliable way to test and cleanup config reloads.

Checklist:

• Test
• Document
• Test on staging cluster for a while