By default this sends out each stat in a single UDP packet. I found
with the AFS monitoring, blasting these out in very quick succession
can result in stats getting lost fairly easily.
statsd's pipelines are really designed for this. It batches up all
the stats and then sends them out combined in reasonable MTU sized
chunks.
Change-Id: Ife520d549da3a1a667be15e95a747e313825ac20
d514389e41