[ML] excludeDeleting option of JobConfigProvider.expandJobs and JobConfigProvider.expandJobsIds leads to unintuitive error messages

[droberts195]

The JobConfigProvider.expandJobs and JobConfigProvider.expandJobIds methods have an excludeDeleting argument that excludes deleting jobs from the returned list.

This method is then used in most of our endpoints that accept wildcarded job ID patterns, so that, for example, you cannot close a job that is being deleted.

However, the downside of the way this works is that if you call one of these endpoints against a job that is being deleted you will be told that the job doesn't exist. This is extremely confusing.

For example, if you try to close a job that is being deleted, you'll receive a resource not found exception. This is inconsistent with the decision we made that closing a closed job is not an error. In the case of a job whose deletion is proving problematic for some other reason it also makes troubleshooting harder, because the output from one request may lead the people doing the troubleshooting to think the job no longer exists (i.e. the desired state) when it still does exist but is stuck in the deleting state.

[elasticmachine]

Pinging @elastic/ml-core (Team:ML)