Refine iRT autorefinement for large libraries.

Currently with large in-silico libraries, default parameters for auto iRT refinement fails due to the low proportion of true signals present in the data. For this case it might be worth looking into a spectrum-centric apporach for finding the calibrant peptides. 