fixes Bug 853299: added statsd to legacy processor

[twobraids]

This PR adds statsd support to the new processor. It was evidently not updated when statsd was added to the old processor. This oversight is corrected.

I added a statsd wrapper class that adds configman support to the StatsClient class. It offers a dynamic loading of statsd and the ability to turn counters on and off in configuration.

[lonnen]

Re-ran this in leeroy because the failure seemed unrelated to the patch, but the same failure reoccured (see builds 600, 601). It is not affecting other PRs, or socorro-master.

[lonnen]

@twobraids what was the problem earlier?

[twobraids]

I wish I knew. I started a new branch based on a later master and introduced my changes one at a time. The sum of the parts passed. Then I rebased to one commit and that also passed. No clue as to what was the actual cause of the problem.

[lonnen]

empty doc string?

[lonnen]

@twobraids ok. Happy to just move on from that.

I'm not sure about the empty docstring and did you consider metrics instead of statistics as the module name? Otherwise this is a-ok with me.

[lonnen]

One more thought: single segments of a metric that may contain dots, like hostnames, should have the dots replaced with _. This can be done by sanitizing risky keys, or by changing the stats module to accept lists of strings, sanitizing the list, and '.'join(that_list). I'm thinking the later is overkill.

Come to think of it, nothing jumps out at me here as a risky key... maybe the Processor name? Ignore this bit if I've described a problem we aren't having.

[twobraids]

I think the method currently in use is a faithful reproduction of the method currently in use in the production processor.