Upgrade assistant never finishes loading reindex button if `.watches` is a concrete index

[gwbrown]

Kibana version: 6.7.0-SNAPSHOT

Elasticsearch version: 6.7.0-SNAPSHOT

Server OS version: Docker

Browser version: Firefox 65.0.1

Browser OS version: MacOS 10.14.2

Original install method (e.g. download page, yum, from source, etc.): Docker

Describe the bug: If .watches is a concrete index, rather than an alias, the Reindex button will never finish loading. This is due to the fact that trying to GET .watches returns an error.

This can happen if a cluster that uses Watches was created as a 5.x cluster, then upgraded to 6.x without running the Reindex Helper (or Migration Upgrade API) in 5.x before upgrading.

Steps to reproduce:

1. Create a 5.x cluster which uses Watcher.
2. Upgrade to 6.7 without reindexing Watcher via the 5.6 Reindex Helper (or Migration Upgrade API)
3. Load the 6.7 Migration Assistant

Expected behavior: The button should either work as expected, or at least show an error message with a description of what to do

Screenshots (if relevant):

Any additional context: The Migration Upgrade API is still present in 6.7 - although it's not used for the 6->7 upgrade, it can still be used to "fix" this situation, where it should have been run before upgrading the cluster to 6.x. Running POST /_xpack/migration/upgrade/.watches fixes the issue.

[elasticmachine]

Pinging @elastic/kibana-operations

[gwbrown]

This should no longer appear to users now that elastic/elasticsearch#39609 has been merged, so I'm going to close out this issue.