Updates for CP5 tune

Main changes

• Update weights for use with samples produced with the CP5 tune, and for newly tuned Bowler-Lund or Peterson parameters assuming the CP5 tune (following dicsussions with @mseidel). The tuning of the fragmenation parameters itself was updated here: https://gitlab.cern.ch/cms-gen/Tuning/-/merge_requests/20
• Provide both jet pt-averaged (as before) and jet pt-dependent weights (i.e. 2D in xb and gen b jet pt), to account for the observed pt dependence
• Fix tagging of semileptonic B hadron decays, for computing branching ratio systematics: use "tag" decay leptons, now included in the ParticleLevelProducer (from 10_6_19_patch2 on), instead of simply relying on the presence of leptons clustered inside the jet. This now properly recovers the Pythia8 branching ratios (were over-estimated by about 2x before).

More technical changes

• Use leading B hadron in the jet: before, the leading hadron (regardless of flavour) was taken, and the weight computed only if that hadron was a B hadron. The results are slightly different for low xb values (i.e. in the cases where another non-B hadron with higher pT is present).
• Change weight smoothing mechanism from ROOT TSpline2 to Akima sub-spline (numpy)
• Add some scripts for producing validation plots and update those for running on condor
• Compute distributions of xb by generating ttbar events on the fly based on powheg gridpack, instead of using existing LHE events (since they are not stored anymore in central production)
• Run clang format

Note that this PR includes the changes from the 94X branch, which I would suggest should now go into the master branch. Perhaps a "legacy" 80X branch can be kept, forked from the current master branch?

The results will be discussed at the TMG meeting next Monday: https://indico.cern.ch/event/1004752/