For an index rebuild, the workflow service will load all workflows into
memory at once. Depending on the amount of workflows in the system, this
can cause a huge, sudden spike in memory usage which may even cause the
whole process to fail. E.g. on one larger system we managed to kill the
process on a machine with 64GB memoty…
This patch introduces pagination for the workflows. This means that only
a limited number of workflows are loaded from the database and processed
at the same time.