Add registry to track status

[phillbaker]

So that the beat can be run in a way that is not dependent on one machine, we should track the status of each metric's scraping in a registry. This could also be used to backfill data on the first run if no previous timestamp is present.

This could be similar to https://github.com/e-travel/cloudwatchlogsbeat/blob/fe67ee1cfe06e08cc1338ceb1417efc32efa5d9a/beater/s3.go