add monotonic_count support

[arbll]

What does this PR do

Adds the ability to jmxfetch to consume "monotonic_counts". This new metric type for jmxfetch is going to be almost identical to the agent checks monotonic_count:

Submit the sampled raw value of your counter. Don’t normalize the values to a rate, or calculate the deltas before submitting. If the value of your counter ever decreases between submissions the resulting stored value for that submission is 0
https://docs.datadoghq.com/developers/metrics/counts/

The only difference with agent checks is that the value will be stored as a rate in the backend. Functionality wise this should not have any impact.

This will allow jmxfetch to monitor counters that increase over time like the total number of request received by a web server and most importantly store them as counts in Datadog.

Notes

This is based on a POC from @mirkoprescha that you can find here: #190.

[olivielpeau]

Looks good overall, thanks!

If there are still some unknowns about how the in-app metric type will behave, please list them here.

Also, we should add some documentation about this new JMXFetch metric type. Do you know where it'd be best to document it? (cc @l0k0ms )

[l0k0ms]

I think we should update the java integration here: https://docs.datadoghq.com/integrations/java/#the-attribute-filter. Just open an issue on the doc github repo or in our trello board for tracking and we will do it.

[truthbk]

Force-pushed after rebasing on top of master following the merge of #98, also fixed a couple of stylistic issues.

[olivielpeau]

(and please open a request for a doc update ;) )