Fixes the command that changes Shared Runners timeout.

For the migration process we need to ensure, that during the switch
moment no job on Shared Runners will be running. The baclout window
was scheduled to take 1 hour, and this is the timeout that we want
to enforce on Runners.

While we need to update the general Shared Runners named
`shared-runners-manager-X.gitlab.com` - since the timeout is currently
set to 3 hours, we should not touch `gitlab-org` labelled Shared
Runners named `gitlab-shared-runners-manager-X.gitlab.com`.

`gitlab-org` runners already are configured with a 1 hour timeout
and this is the expected timeout for them.

Since it aligns with what we need for the migration, I've updated
the commands that are changeing the timeout. Now we're not touching
the `gitlab-org` runners at all, which means that:

- there is no change during the preparation step (`gitlab-org` runners
  during the migration will be set to 1h timeout),
- the reconfiguration step (or the failover step) will not change
  the expected timeout of `gitlab-org` runners, which should be left
  as 1h after the migration is ended.
2 jobs for fix-runners-timeout-command in 1 minute and 1 second (queued for 4 seconds)
Status Job ID Name Coverage
  Test
passed #2711880
references

00:01:00

passed #2711879
shellcheck

00:01:01