Final fix for best_track

This issue is the last episode of the successful series: how to chose the best track (shower) of a given reconstruction algorithm from a KM3NeT ROOT file.

The final approach is:

• provide a function with e.g. best_track(tracks, rec_stages) and the user has to provide the rec_stages. It can be a single one, a list or a range etc., always using the named definitions from km3io
• whatever stages are selected, to pick the "best track", the function will always take the one with the longest rec_stage vector and highest value of the likelihood parameter after all other cuts have been done
• optional "hardcoded" functions to get the best track, so be compatible with what's available in Jpp, resp.km3net-dataformat/tools/reconstruction.hh (best_jmuon, best_jshower, best_aashower, best_dusjshower).

Fin.

What do you think?