Pinging @elastic/ml-core (:ml)

This was dropped during the backport indeed. But the reason is different. In 7.x, we need to register named writeables via XPackClientPlugin.getNamedWriteables() in order to keep working with the transport client. The latter is removed in master so we have refactored the way we register named writeables there in #53757. However, this discrepancy between master and 7.x can cause backport issues like this one. For consistency, could you register the autoscaling writeables in XPackClientPlugin in 7.x?

Given that autoscaling is still experimental, I think we should not move things to xpack-core ahead of time. I have put it on our autoscaling sync for next week to discuss this. I would be inclined to just document that autoscaling does not work with the transport client.

Moving the specific ML class would require moving other parts of autoscaling to xpack-core too. And for it to be functional, actions, requests and responses have to move too.

So for now, I would prefer to get this PR in as is to enable consumers of the API to start testing ML autoscaling and then we can tackle the move to xpack core (if that is the decision taken) in a follow-up. Would that be ok?

Thanks Dimitris!