Should virtual host-specific definition import exclude virtual host name?

[michaelklishin]

See #10068 for the background.

During that investigation we have learnt that in early 2016 in 8ba0e07 a decision was to exclude the vhost field when exporting definitions of a single virtual host. I don't remember the details but the goal
could have been to allow for such files to be imported into any virtual host in a different cluster.

We can restore the vhost field to avoid the confusing scenario in #10068 and possibly ignore it during the import operation.

[michaelklishin]

Discussed this one more time as part of my work on #12821. This is something we can change for 4.1.0.

[michaelklishin]

We have made a decision. Per discussion with @mkuratczyk:

1. Being able to import a single virtual host into another one in a different cluster and with a different name has its benefits
2. We can add some metadata to the definitions to help determine the definition file "type" (cluster-wide definitions vs. single virtual host definitions)
3. We can produce a better error message on boot

and keep everything as is, plus backwards compatible for existing deployment and practices, which is also very valuable.