Collect runtime metrics from Ruby environment

[delner]

In an effort to improve insight into application performance, we'd like to collect metrics that describe the Ruby environment, so that this data can be displayed with traces in the Datadog trace page.

This pull request adds two major new parts to the tracer:

• Datadog::Metrics which is an application level Statsd instance which can be used to send metrics. If Statsd is available, it will autoconfigure its client to send stats to the agent on UDP port 8125.
• Datadog::Runtime namespace which contains functions for collecting stats from the runtime environment. This is used in the AsyncWorker thread to collect and broadcast runtime metrics to the client configured withinDatadog::Metrics.

TODO:

• Write unit tests
• Add runtime-id tag
• Add service tag

[brettlangdon]

cool, no major issues I see

[pawelchcki]

Great Job @delner! Looks like really awesome addition!

Left one suggestion about using counters to correctly process some metrics. Otherwise looks good!

[brettlangdon]

one minor doc change, otherwise this is looking great

We are deferring requested changes since it doesn't really effect Ruby metrics at this time.