Follow-up from LHCb!2257
We should store more track-pv information from the PV reconstruction. The "weights" alone aren't overly interesting, one also needs the derivatives and IPchi^2. Both are available within the PV reconstruction and "only" need to be stored (and fiddled through a bunch of interfaces). This should be possible regardless of which way PV-track association is stored.
I'm assigning this to myself now as I have had my fingers on the Iterable PVs anyway, though I'll need some help from the algorithms authors about where to find the numbers in the algorithm implementations.