masking is performance critical
masking has a very high impact of performance. All data is copied, altered in masking, and every column is again modified with unmasking, wich makes saqc very slow for big datasets.
Thus i suggest to reconsider to move masking in the flagFunctions itself, because every function knows which data to read and write and so can limit the masking to those columns.
i improved the masking somewhat in the perf_test [1] branch. but still its to slow IMO.