Improve performance for publish/discard of workspaces by using the synchronous mode

[skurfuerst]

During publishing of nodes, we usually fork sub-processes for event processing (for every event). This drastically slows down these operations. We can enable synchronous mode in these methods, to speed up the system at least by a factor 2-3 for these operations.

It could have some side-effects if caches were not properly cleared,
but at least I did not see any yet (and we run tests often with this flag).

Related: #4388