Create a MetricsReporter dedicated for Stream introspection

Create a custom MetricsReporter that is going to filter existing metrics:

• computation failures
• computation active threads
• global streams/consumers positions (pos, end, lag, and latency) this will be reported only by one node in the cluster
• JVM CPU usage
• JVM mem pressure

Metrics will be serialized in a stream with a single record with additional metadata:

• node id
• node IP

Using this dedicated metrics reporter removes the need for accessing the existing monitoring system (DD) from Nuxeo.