Improved robustness of the Velo Tracking

The Velo tracking algorithm was developed based on the idea described in the TP for Level-1 trigger, namely to start by a triplet of consecutive stations. During the recent LHCC review, this was pointed out as being quite sensitive to detector inefficiencies, for example dead strips, which were not in the simulation. Since then, I have implemented a small modification to the Velo tracking algorithm, which reduces to an acceptable level the sensitivity to detector inefficiencies. This short note describes the original situation, the changes and the new performance, with some comments on possible future improvements.