introduce weight cache for performance
Description
Rather than working out the process weight for every event, this change set does this once for the first event, then uses a weight cache to store the result and does a cheap look-up for subsequent events. This results in a ~7% speed-up for the ATLAS V+jets setup (on top of all the other performance improvements from the last couple of years). Seems too trivial not have.
Test Plan
Validated by Ilektra on Stefan's ME branch for a few phase-space points.
Additional Checklist Items
-
I have updated documentation as needed -
I considered if this should be cherry-picked/ported somewhere else (rel-2-x or master), and if yes, opened a corresponding issue or MR