ClearML health check failing with autoscalar only

[johnml1135]

When there is no workers available on a queue, Serval complains that it is "unhealthy", but this is the normal state for the auto-scalar. We should be able to detect that the autoscalar is listening to this queue and therefore say "it's ok - don't freak out".

[johnml1135]

We currently use the https://clear.ml/docs/latest/docs/references/api/workers/ endpoint, but there is no corresponding one for the autoscalar, at least what I can see right now.