fix Velo track duplication (in persistreco)

Needs Rec!4085 (merged)

Fixes bug of duplication of velo tracks. Sort of alternative to !3781 (closed). Saves about 3% of HLT2 bandwidth (tested on MC).

Some notes on persistency and why copying of tracks in the first place: to disentangle the quite complicated and different variants of reconstruction from what tracks to persist and how this is done, as PyConf doesn't allow a posteriori changing of output locations (yet). Intermediate solution is to change this to selections (only 'pointers' in the persisted locations), this was always the plan, but requires persistreco version change, this is at least forseen for 2025.

Impact on (un)packing is minimal and behind the scenes and is save (as discussed offline with @graven). Changing the physical location of the Velo tracks for the PV tracks persistency is possible/foreseen by the current persistency.