Metrics Output

[jantman]

My employer now has need for a built-in method of outputting metrics - per-limit usage and current limit, as well as overall runtime - to a time-series service. We currently use Datadog. I'm being allowed to implement this on work time.

We're currently doing this via a Python script that wraps awslimitchecker's Python API, similar to the suggestions I made in #152 (statsd/graphite) and #256 (prometheus), the latter of which resulted in merging docs/examples/prometheus.py.

My plan is to implement a base MetricSink class with a relatively simple interface (likely similar to what I described in this comment), documentation for implementing subclasses to support specific metrics stores, and the first subclass for Datadog.

I think the only undecided bit so far is how to handle third-party dependencies - do we make them extras and just have other metrics subclasses merged into the project itself, or do we distribute the plugins as separate packages and use setuptools entrypoints for plugin discovery?

[jantman]

This was merged to master in #427 and released as 7.1.0. The release is now live on PyPI and the Docker image should be live on Docker Hub shortly.