Limit the number of passing combinations in ParticleCombiner

During the HLT2 reprocessing of 2022 data some jobs on the grid failed, essentially due to nodes running out of memory. The underlying cause were too large combinatorics in some combiners, such that the DstData bank would have been flooded with particle- and vertex-combinations. This MR limits the number of successful[1] combinations to 2000[2].

There are several things that will need to be discussed before moving on with this.

[1] passing all combination- and composite cuts as well as the vertex fit
[2] This is a random but large enough number that I think was used in Run2 somewhere