Fix state sets among Pr <-> v1 <-> v3 conversion (+ Add filter on CaloMatch relations)

This is best reviewed commit-wise:

1. implements a filter consuming a RelationTable the from relation of which are tracks new feature

2. templates the PrMatchNN such that it can take v3 seed tracks as input (e.g. the output from the relation filter added in 1.)

3. contains the majority of changed files and simplifies/adapts code following LHCb!4429's available state locations bug fix

   Click to expand

   • fixes the conversion from PrSeedTrack -> v1 confusing the EndT state with LastMeasurement (this is causing Calo ref changes!)
   • fixes the conversion from v1 Downstream -> v3 Downstream which ignored the BegRich1 and EndRich1 and probably caused undefined behaviour
   • fixes states on PrUpstreamTracks: now the tracks actually have a state in the UT which can and is used by the PrKF avoiding unnecessary extrapolation
   • fixes loss of information of Ttrack v1 -> v3 conversion

4. adds two functors working together with the relation filter, namely GET and MIN_ELEMENT_NOTZERO

Goes with LHCb!4429 and Allen!1433 and Moore!3045

Throughput is not affected 👍 see HLT2 MR vs. HLT2 master and HLT1(CPU) MR HLT1(CPU) master

Will close #519 and is related to Moore#657