Reorganise track fit implementations into separate classes

I reorganized the implementation of the track fit for the 3 cases (x-vs-z, y-vs-z and xy-vs-z) of the Hybrid Seeding. There is no difference in the physics performance, i.e. the output of the Hybrid Seeding is identical to the previous version. Thanks to a better optmized code, the CPU time goes down by about 2%.

The main reason for these changes is to make the code better organized and easier to maintain.

References updated in Brunel!1001 (merged) and Moore!467 (merged)