trackingLST modifier to enable LST via cmsDriver and trackingIters01 to run just the leading two iterations

[slava77]

• trackingLST modifier and its initial use to enable LST as a replacement to the highPt triplet step building
  • the initialStep tracks are only virtual (not used in the generalTracks); the initialStep cluster masks are used only in highPtTriplet step seeding
  • eventually this setup can be combined/compared with pixel tracks from patatrack replacing seeding from the initial and highPtTriplet steps
• added SeedStopInfo output to the LSTOutputConverter mainly to keep DQM running; the info is currently dummy (an empty vector)

Can be tested with --procModifiers gpu,trackingLST combined with trackingOnly setup.

Here are some results with PU for allTracks (aka generalTracks) for 100 events in ttbar without PU

all plots are in http://uaf-10.t2.ucsd.edu/~slava77/figs/lst/CMSSW_13_0_0_pre4-lst12/mtv-oob-trackingLST-rmIS-mask-8c1d76e/

The comparisons are for:

• default phase-2 tracking (blue)
• LST replacing the highPtTriplet while initialStep tracks are still used (red)
  • the "to be virtualized" initialStep tracks contribute redundantly, efficiency is high
• as above, with initialStep tracks removed from merged generalTracks collection (black)
  • only LST is now covering for what was initialStep and highPtTriplet, BUT the cluster masks are still as if the initialStep was there; this does not allow the following iterations to recover
• LST replacing the highPtTriplet and initialStep including cluster masks (orange)
  • the efficiency now recovers to roughly the baseline
  • the fake rate is about the same
  • the duplicate rate goes up significantly, primarily in endcaps where the pixel (IT) part of LST is using only 3 or 4 layers out of more possibly available: these are found again by later iterations.

TODO:

• tune duplicate merging
• ~make a tracking configuration with just the initialStep and highPtTriplet iterations as a reference~ (done in the last commit )

[slava77]

http://uaf-10.t2.ucsd.edu/~slava77/figs/lst/CMSSW_13_0_0_pre4-lst12/mtv-oob-trackingLST-rmIS-mask-8c1d76e-2 has just two inputs: default phase2 and this PR as of 8c1d76e with LST

[slava77]

the last commit 03fac7e now has trackingIters01, which can disable all the recovery iterations of iterative tracking (phase-2 setup only) and leaving just the initial and highPtTriplet steps.

more plots in http://uaf-10.t2.ucsd.edu/~slava77/figs/lst/CMSSW_13_0_0_pre4-lst12/mtv-oob-trackingLST-Iters01-03fac7e/plots_ootb.html