Allow access to some plugins when es plugin is red

[ppf2]

Starting in ES 2.0, we have moved Sense into a Kibana app. One use case that arises is that you can no longer access the Sense editor when the .kibana index has not completed initialization:

In this particular case, the .kibana index has shards that cannot be started up because it is failing disk allocation threshold checks, eg.

[2015-11-03 16:54:46,233][WARN ][cluster.routing.allocation.decider] [node1] high disk watermark [90%] exceeded on [t3miuXlFQLSnoNkUdezszQ][node1][path_to_problem_shard] free: 11.9gb[5.1%], shards will be relocated away from this node

Hence:

index                 shard prirep state      docs   store ip        node  
.kibana               0     p      UNASSIGNED                              
.kibana               0     r      UNASSIGNED  

So to address this, the user can either provide more free disk space, or dynamically change the disk allocation settings.

And since the Sense editor does not work until .kibana index is available, they will have to write the command to adjust the disk allocation settings using curl or another REST client.

It will be nice if we will still allow the Sense editor to be accessible, so that:

1. They can run queries to look at things like _cat/shards and update settings in Elasticsearch to resolve the shard initialization issues without having to run these commands from somewhere else.
2. They can use Sense to interact with other (non .kibana) indices.

[msimos]

I also had this problem, also Marvel becomes unavailable (blank screen) if Kibana can't access .marvel-es-data. Because there is a blank screen there's no indication of the issue.

[epixa]

I'm going to close this in favor of #5300