Improve AGIPD stats calculation

Currently the 'bad' pixel-cell (from dynamic masking) propagates though the calculation, i.e. being masked even once it stays at its position within the whole mean_pixelcell. Should be done smarter.

Other possible options to improve:

1. calculate the mean of squared intensity (along with the mean intensity), so that more general approach can be used when filtering [aborted]

2. introduce a parameter, that can switch off dynamical masking at all (needed?)