therefore we don't need the api check.

don't differentiate between the different runner types, instead we rely
on the Runner model to provide the available projects.
scheduling is now applied to all runners equally.

instead of having the explicit logic duplicated from the scope we can
use the scope instead.

the scope `Ci::Runner.belonging_to_group` does not filter out the
runners where the projects has `#group_runners_enabled` set to false
anymore.

it didn't show up in the runners UI anymore when group runners were
disabled. this was confusing.

the flag is only relevant when selecting appropriate runner for a build.

1. we don't want to migrate all existing groups
2. we generate the token when showing the runners page, as this is the
   first time that the token will be used.

as we're using UNION now we can use INNER JOIN.

the unions performs much better than the joins, and the execution time
is constant with a growing number of records.