Draft PR for prototype final fit

[leonardogiannini]

PR description:

The main changes are inside MkFinder.cc where the following functions are implemented

void fwdFitInputTracks(TrackVec &cands, std::vector<int> inds, int beg, int end);

void fwdFitFitTracks(const EventOfHits &eventofhits, const int N_proc, int nFoundHits, std::vector<std::vector<int>> indices_R2Z);`

void bkReFitInputTracks(TrackVec &cands, std::vector<int> inds, int beg, int end);

void bkReFitFitTracks(const EventOfHits &eventofhits, const int N_proc, int nFoundHits, std::vector<std::vector<int>> indices_R2Z);

void reFitOutputTracks(TrackVec &cands, std::vector<int> inds, int beg, int end, int nFoundHits, bool bkw = false);

std::vector<std::vector<int>> reFitIndices(const EventOfHits &eventofhits, const int N_proc, int nFoundHits);

A new function is added to KalmanUtilsMPlex

The function are wrapped inside MkBuilder and the call is made in RecoTracker/MkFitCMS/standalone/buildtestMPlex.cc (commented out for now)

[slava77]

is it practical to start splitting contributions to more than one file?
If we add smoothing and outlier rejection , it's probably another 1K lines if not more