[RFC] Pixel-ClusterLess reconstruction

[VinInn]

DO NOT MERGE:
this PR is mostly in support for the presentation at Pixel DPG on Wed 26/08/2020
https://indico.cern.ch/event/934821/#50-digiless-aka-clusterless-tr

This PR consists in three major modifications of some very basic/fundamental components of CMS tracker/tracking software implementation.

1. make MeasurementDet and some of its ancillary producers (ClusterRemoval) to consume RecHIts instead of Clusters
   this is purely technical as RecHits are just a proxy to the Clusters
   to prepare for next final step, RecHits have been sorted in "x" and is this sorting (and not the cluster sorting by firstRow) that is used in the MeasurementDet: so some regression is expected

2. Make the OmniRef to "reference" an item in a SoA (it is actually very general) and (if this is the case) make GenericCPE to consume the SoA when "improving" the hits using the track angle

Both these changes are fully backward compatible (requires customization of existing HLT menus) and should actually be applied upstream to standard CMSSW

3. add a conditional path in the PixelRecHitFromSoA to make RecHits to point to SoA and not to clusters
   fix code (in particular MeasurementDet and Validation) to work with "cluster-less" hits
   at this point one can drop the conversion to LegacyCluster in patatrack workflows (in case of unscheduled will happen w/o any further intervention)

Please refer to the inlined comments and the presentation referred above for further details.

It also includes code from #528

The way this code will then materialize in actual PRs is open to suggestions and comments
(contrary to most patatrack code it affects mainstream classes belonging mostly to Tracking-POG)