In #769 I made it possible to do filtering of scores during the compute. E.g. if a score is 0 or below a score.

In practice I now we are actually often interested in just getting the top 1 or top 10 scores from a library. This can of course be run after first storing all scores, but to reduce memory load during the compute it would make a lot of sense to apply this during the compute.

Since we already loop over query spectra during the dense matrix compute it can easily be added there. In addition it could be added to the sparse matrix compute as well. But it would require sorting the Mask in a way that all scores are calculated on a row by row basis, making it more complex to implement.