ParticleCombiner: limit the total number of particle combinations
In addition to the limit on composite combinations (2000), this introduces a limit on the number of combinations tried before the vertex fit. The number I chose (50e6) is arbitrary but should be big enough such that even somewhat loose N-body combinations are not killed off but still allows the thread to process the combination in not insane time. I would not go higher with this number, but am very much open to make it smaller. See also #558 (closed)
effect studied in https://lblogbook.cern.ch/HLT/751
Resolves #558 (closed)