Fixes material p2r master

[cerati]

PR description:

Fixes to propagation to R error propagation and to application of material effects (the latter from PR #106)

Note about math derivation:

https://www.overleaf.com/read/rkwnmmgpjzkc

PR validation:

ttbar: http://uaf-10.t2.ucsd.edu/~cerati/pr127_mod/

ttbar (using mkFit for low pt iterations): http://uaf-10.t2.ucsd.edu/~cerati/pr127_lowpt_mod/

mu pt<1 using mkFit only for LowPtQuadStep: http://uaf-10.t2.ucsd.edu/~cerati/pr127_mu_lqonly_mod/

The configuration named mod is the same as this PR but with the following lines commented out: https://github.com/cerati/cmssw/blob/fixes-material-p2r-master/RecoTracker/MkFitCore/src/PropagationMPlex.icc#L171-L175
It looks like mod gives the best results, so somehow the lines above are problematic for some reason.

[leonardogiannini]

0. PR127 (after introducing material changes) http://uaf-10.t2.ucsd.edu/~legianni/plots_compare_127_andMod - should cross check with http://uaf-10.t2.ucsd.edu/~cerati/pr127_mod/
1. checking new windows before 127 (to validate the method) http://uaf-10.t2.ucsd.edu/~legianni/plots_newWindows_0 - (this is also an errata courage wrt to what was shown in the previous internal meeting) - not fully consistent especially in the detached iterations
2. http://uaf-10.t2.ucsd.edu/~legianni/plots_compare_127_newWindows/ - new windows derived with PR 127 but not with the "mod" version. Overall increase in fake rates and small eff. loss - need to repeat in the mod version

[slava77]

http://uaf-10.t2.ucsd.edu/~legianni/plots_compare_127_newWindows/ - new windows derived with PR 127 but not with the "mod" version. Overall increase in fake rates and small eff. loss - need to repeat in the mod version

Thank you for the MTV plots.
We have a pretty large drop in the 0.4 - 0.8 GeV range.
It would be good to get a variant of MTV plots with efficiency plots selected to be below 0.9 GeV to understand the situation better.

Is "original-127" using the old windows from Mario, or is it rederived?
It would be good to compare cases based on the same machinery to decouple the changes in the parameterization/fitting strategy of the windows.
Do you know roughly how the windows change especially in lower pt bins e.g. in the initialStep iteration?
Naively, the dphi/dq windows should not really change, and the chi2 would change. Is that the case?

From the dist plots it looks like the losses are mainly in 0.5 < |eta| < 1.5, but are still significant in the barrel as well. There is something interesting in one bin near eta of 2.5; perhaps some feature in the window selection in the edge of the acceptance?

[leonardogiannini]

This is no longer compatible with master (after the update of the material derivation). We should remember before integration