WIP: Fix bls correction via noise peak
Base-line shift is an effect observed in illuminated images from the AGIPD detector. Normally base-line offset (also called offset correction) is estimated using dark image. However in AGIPD this offset depends on total illumination of a module and shifts to the lower value with large module illumination.
One of the way to correct for this effect is to look at non-illuminated pixels of each image. Distribution of these pixels will build a noise peak. A position of this peak given a value of a base-line shift.
In order to be able to estimate a noise peak image should have enough non-illuminated pixels. This MR introduces a threshold on upper value of fraction of high-gain pixels below 50 AGU.
Another change is no fix bin-side of a histogram to be about 1 ADU. This helps, when I run on flat-field data.
This MR should be tested on different data samples: Low illuminated flat-field data, high uniformly-illuminated data, high non-uniformly-illuminated data.