Lumi counters: loop only over real velo clusters
Currently, the "# of velo clusters" lumi counters loop over all Clusters between the offset for the first module pair in that event and the offset of the first module pair in the following event. However, the Clusters appear to not be densely packed so erroneous Clusters may exist at end of the list for each module pair.
This MR instead only considers the number of clusters stored in dev_module_cluster_num for each module pair.
From tests, it seems the first out-of-range Cluster always has the correct LHCbIDType (i.e. VELO), implying it could be real. I'm not sure why this is the case but I'm trusting that there isn't an out-by-one error in dev_module_cluster_num (if there were, it would make it impossible for a module pair to have zero clusters as dev_module_cluster_num is an unsigned, so this seems reasonable).
Addresses #462