Fix polydisperse 2DGRM with nBound=0 for some particle type

In CADET-Verification, i found that for a polydisperse (NPARTYPE=4) 2DGRM where one particle type has "NBOUND=0", simulation results are different between current master and the last release.
I don't see an obvious reason on first glance but that bug must've snuk in during #471 .
An MRE is easily created by running the respective benchmark in cadet-verification