load T5 TCs and make seeds for them on the fly

[slava77]

T5s are added

• the seed (TrajectorySeed) is made using SeedFromConsecutiveHitsCreator (all 10 hits at this point); the seed state is defined on the last hit.
• the TrackCandidate state is just a propagation of the seed to the first hit
• most of the code was already there, it just needed a few relatively minor cleanups.

As of e2cff6df: on 100 events I observe that only about 96% of T5s are converted to TrackCandidate. Most of the inefficiency is due to failed seed making with a minor additional loss from fairly poor seed state parameters

all preliminary plots (as of e2cff6df; READ TO THE END to see the final performance) are in http://uaf-10.t2.ucsd.edu/~slava77/figs/lst/CMSSW_13_0_0_pre4-lst12/mtv-oob-trackingLST-Iters01-T5-e2cff6d/plots_ootb.html this is for ttbar WITHOUT PU (100 events)

black -> orange is the impact of this PR on tracking with LST as of e2cff6df.

---

As of a16ef33: on 100 events all T5s are converted to TrackCandidate and the efficiency is up wrt e2cff6df by a couple %. In this commit (33e94a6), the requirement that the first two hits are on P layers is added. The SeedFromConsecutiveHitsCreator relies on the first two hits to make the initial kinematics estimate.

all plots are in http://uaf-10.t2.ucsd.edu/~slava77/figs/lst/CMSSW_13_0_0_pre4-lst12/mtv-oob-trackingLST-Iters01-T5-2P-a16ef33/

black -> orange is the impact of this PR on tracking with LST as of a16ef33 (the last commit).

[slava77]

As of 33e94a6: on 100 events all T5s are converted to TrackCandidate and the efficiency is up wrt e2cff6df by a couple %. In this commit (33e94a6), the requirement that the first two hits are on P layers is added. The SeedFromConsecutiveHitsCreator relies on the first two hits to make the initial kinematics estimate.

all plots are in http://uaf-10.t2.ucsd.edu/~slava77/figs/lst/CMSSW_13_0_0_pre4-lst12/mtv-oob-trackingLST-Iters01-T5-2P-33e94a6/

black -> orange is the impact of this PR on tracking with LST as of 33e94a6 .

~(this is added in the PR description as well)~

[slava77]

I found that some T5 seeds were truncated due to a wrong condition; this was fixed in a16ef33. The impact on efficiency is rather minimal

[slava77]

I made some quick analysis, just counting objects in TrackCandidates (output of LSTOutputConverter) and in the reco::Tracks (built tracks, after the final fit) and I see that there is 0.6% loss. This would eventually need to be understood, but not short term (not now).

So, I think that the PR is ready for merging. @VourMa