FPGATrackSim: fix to maxBranches calculation in window extension tool
The first version of this was trying to recompute phi and z distances, but we already computed them to actually do the window cuts.
This version uses the phi and z distances as computed for the window algorithm and saves a tuple of (dphi, dz, hit). It then sorts the vector of these tuples, and then uses the lowest ones to populate the road.
I think the practical difference is that when I recomputed these I was using the average radius of the "layer" of the region instead of the actual radius of the hit, which would have an outsize impact on the forward region. With this the F-600 ttbar efficiency seems to stay a lot more stable in eta.