Preselection based on peak-to-peak analysis

[paulmueller]

In dcnum <= 0.14.0, preselection was employed to remove input images that presumably contain no events. This was implemented as follows:

if self.preselect:
    ptp = np.ptp(self.data.image_corr[index])
    if ptp < 0.1 * self.ptp_median:
        # do not use this event
        return None

The ptp_median value was computed from the background data.

Preselection was removed in dcnum 0.15.0, because

• it represents a preprocessing step that has possibly hidden impact on data processing
• it is not yet represented in any pipeline identifier

Should we want to re-introduce preselection, then we would need to properly at it with its own pipeline identifier. Should the input data contain duplicate frames, this could also be addressed (instead of segmenting a duplicate frame twice).