[ML] Saved object syncing can throw an error with datafeed ID update

[jgowdyelastic]

When syncing or initialising AD saved objects, if a job which lives in a different space is missing the datafeed ID but the datafeed exists, syncing will throw an error.

To reproduce:

• Create a new space (space2)
• PUT a job using the es api
• In space2, Run sync in the UI so that the saved object for this job is created, but without a datafeed ID
• PUT the datafeed using the es api
• In the default space, open the sync flyout and see that it wants to fix this new job to add the datafeed ID.
• Running sync throws an error which is shown in a toast.

This also affects the hourly auto sync task which only runs on the default space. Meaning it's not possible to update the datafeed ID for jobs that live exclusively in a non default space.

This is not a non-trivial fix as the saved objects client does not allow editing of SOs from other spaces. The is an old feature request to allow this

The work around is to run sync in the space which the jobs exist.

The datafeed ID updating code should catch the error and not cause the endpoint to return with a 500 error code.

[elasticmachine]

Pinging @elastic/ml-ui (:ml)