Atlas Cloud Watch: Add the CloudWatchMetricsProcessor, an abstract cl…

…ass (#413)

for writing and reading from a backing store implementation. The processor
is responsible for matching incoming fire hose data to rules and merging
new values with existing values in the cache. It is also responsible
for regularly scraping the cache and publishing the final data.
Included is a simple local hash map based store for unit testing.

atlas-cloudwatch/src/main/resources/reference.conf

@@ -102,6 +102,16 @@ atlas {
       }
     }
 
+    // TEMP: Used while testing cloud watch streaming to prepend "TEST." to metrics in order to compare against
+    // the poller.
+    testMode = true
+
+    // How often to run the publish thread.
+    step = 60s
+
+    // How long to wait after the top of the step (usually minute) before starting the scrape.
+    publishOffset = 5s
+
     // Batch size for flushing data back to the poller manager
     batch-size = 1000