This PR teaches the `elasticsearch/ccr` metricset to index documents into `.monitoring-es-6-*` indices when `xpack.enabled: true` is set in `modules.d/elasticsearch.yml`.

### To test this PR
The idea is that metricbeat (specifically the elasticsearch/ccr metricset with `xpack.enabled: true`) will create exactly the same documents in `.monitoring-es-*` indices as Elasticsearch's internal collection and reporting does today. 

1. Start up Elasticsearch (using the latest build from `master`).
2. Start up Kibana.
3. Start a trial license from the Kibana Management UI.
3. Enable Monitoring in Elasticsearch (via the cluster setting `xpack.monitoring.collection.enabled: true`). You can do this via [Elasticsearch's Cluster Update Settings API](https://www.elastic.co/guide/en/elasticsearch/reference/current/cluster-update-settings.html) or by clicking the "Turn on Monitoring" button in the Monitoring UI in Kibana.
4. Set up CCR with a leader and follower index.
4. Let Elasticsearch run for ~20 seconds so a few documents are indexed into `.monitoring-es-6-*`.
6. From `.monitoring-es-6-*`, retrieve a document for `type = ccr_stats`
8. Turn off Elasticsearch's internal collection via the cluster setting `xpack.monitoring.elasticsearch.collection.enabled: false`.
7. Delete `.monitoring-es-6-*` indices.
10. Enable the `elasticsearch` module in metricbeat: `./metricbeat modules enable elasticsearch`.
11. In `modules.d/elasticsearch.yml`, add the `ccr` metricset and set `xpack.enabled: true`.  Concretely, your `modules.d/elasticsearch.yml` should look something like this:

      ```yaml
      - module: elasticsearch
         metricsets:
         - ccr
         period: 10s
         hosts: ["http://localhost:9200"]
         #username: "user"
         #password: "secret"
         xpack.enabled: true
       ```


12. Start metricbeat.
13. Let metricbeat run for ~20 seconds so a few documents are indexed into `.monitoring-es-*`.
14. Stop metricbeat
6. From `.monitoring-es-6-*`, retrieve a document for `type = ccr_stats`
16. Using a tool such as http://www.jsondiff.com, compare the documents indexed by Elasticsearch with those indexed by metricbeat. Verify that their structures are identical (same fields, not necessarily same values), except for these known and expected differences:
    1. Only Metricbeat-indexed documents are expected to contain the fields `@timestamp`, `beat`, `host`, and `metricset`. These are "standard" fields added by beats and metricbeat and don't have an adverse impact since they are additive.
    3. Only Elasticsearch-indexed documents are expected to contain the field `source_node`. This field is used for debugging purposes only and not actually consumed by either the Monitoring UI or Telemetry feature in Kibana.


(cherry picked from commit 35068ee)