[CI] add split comparison for ltv

This adds a new comparison step to the LTV jobs. Additionally to the comparison of the whole output files, the files are split up back into their original pieces and are compared piece-by-piece. This should give additional insights into which parts for the long files are still problematic. Another thing we hope for is that the overall value distribution looks a bit better when looking at the split sections as e.g. for MLD, the worst case over the whole file is reported, so if only one segment of the file would be bad, but the rest is fine, the overall picture should look different than now.

This new data is generated additionally to the old comparison outputs, i.e. the whole-file comparison is still in there and also the histograms are created for both whole-file and split comparison separately. So this does not change anything of the old mechanism, but just adds a new thing.