move to a manual tarball of statsd, as github dynamic tgz gen seems broken, prevents npm install from working

some rpmbuild fixes

upgrade to a branch of statsd that lazily closes ephemeral UDP ports, so that we're not allocating one UDP socket per request when under load.  closes #680

issue #704

close #710

(loadgen) fix subtle issue whereby a XXX@loadgen.domain users's password could be changed during load generation, and affect subsequent runs.  closes #681

(loadgen) fix bug in reauth activity.  in addition to clearing cookies, we must clear client state to cause re-authentication to occur.  maybe the cause for issue #682

close #685

close #690

IE8 does not support delete localStorage.[itemname], so this has to be changed to localStorage.removeItem(itemname)

close #688

* Reducing the number of setTimeouts used.
* Using completion callbacks instead.
* Adding a cancelEvent to the helpers which is a decorator around DOM handlers to cancel the event.

close #686

close #685

update to latest jschannel - closes #534 (innocuous console errors when messages are posted to the frame where include.js is loaded that do not have valid JSON as a payload)

run mysql tests *first* (they're more important), and if we can connect to db but browserid table doesn't exist, we still have db connectivity and can run tests

remove process.exit() from loadgen upon email addition failure.  loadgen shouldn't stop till it reaches the top.

disable http agent in http_forward - it has been capping us at five simul forwarded requests per backend process (dbwriter, keysigner)