The begin and end events of a remote refactoring job are handled on different threads, which breaks the observation context

This is actually a much broader problem impacting also extensions jobs that I started thinking of for XCOMMONS-3251. We need in general to have a better clustering handling of jobs, and events produced by jobs: we need a way to indicate jobs which needs to be taken into account by other cluster members, especially grouped jobs to deal with forbidden concurrent modifications (refactoring, extensions, etc.) and remote events need to be injected through those placeholder jobs threads.