cueball HTTP agent config missing pinger or TCP keepalives

[arekinath]

The current CueballHttpAgent config we're using in cloudapi is missing a pinger or TCP keepalive settings, which means that we can easily get "zombie" connections or races when cloudapi is talking to other components (like imgapi or napi) -- while cloudapi is writing a request out to a socket, the other end is closing it, and we get ECONNRESET errors.

It seems like all of the downstream services have a /ping endpoint, or at least don't respond with an error to that URI, so we should just need to add:

        "tcpKeepAliveInitialDelay": 10000,
        "ping": "/ping",
        "pingInterval": 60000,

We'll have to update the version of cueball as well, since the version we're using at the moment (2.1.1) has a bug that causes pingers to misbehave (TritonDataCenter/node-cueball#102)