Monitoring UI inaccessible if remote CCS cluster is down

[n0othing]

Kibana version: tested on version 7.6.0

Elasticsearch version: 7.6.0

Describe the bug:

The Stack Monitoring UI becomes inaccessible if a remote CCS cluster becomes disconnected. When this happens, Monitoring documents are still being written to the .monitoring-es-7-* index.

Steps to reproduce:

1. Start two separate clusters (e.g cluster-1 and cluster-2)
2. Connect Kibana to cluster-1 and enable Monitoring.
3. Connect cluster-2 to cluster-1 as a remote CCS cluster.
4. Shut down cluster-2.
5. Try navigating/refreshing Stack Monitoring

Screenshots (if relevant):

[elasticmachine]

Pinging @elastic/stack-monitoring (Team:Monitoring)

[pickypg]

Workaround:

# 6.1 - 7.6
xpack.monitoring.ccs.enabled: false

# 7.7+ (looking at the repo)
monitoring.ui.ccs.enabled: false

I was going to make a docs PR, but I don't have time to deal with backporting and the setting change:

Set to false to disable the usage of Cross Cluster Search for the Monitoring UI in Kibana. This largely impacts the landing page for the Monitoring UI, which attempts to discover Monitoring data using the local and all (*) remote clusters. After selecting any monitored cluster, the UI will intelligently reuse the same cluster (remote or otherwise) to find it. Defaults to true.

---

Because of how this query is only sent on the landing page, the experience could be improved a little by introspecting the error returned by Elasticsearch and, if it can be detected that it relates to CCS, simply send another request without CCS and use whatever that returns (plus returning the error, which is something that would be new functionality -- data plus an error).