WIP: Fast threshold tuning using binary search and S-Curve fit enhancements

• Fit-able range finding of noisy S-Curve data is finally stable and almost as good as a human in range classification
• Add fast global and local threshold tuning using binary search.
• Global tuning works for all flavours, but not at the same time. Stable even for low thresholds where many pixel are already noisy. Tuning loop takes 1 min.
• Local tuning works for linear flavour with reasonable results. Takes about 1m:10s for linear flavour.

TODO:

• Add differential flavour for local tuning
• Fix untuned pixels that are stuck
• Demonstrate equal performance to existing meta script with slow threshold scan based tuning
• Add very simple "meta" tuning script for threshold (global + local) Does not make sense with the actual restructuring of the code base