Dynamic scoring minor update

[lisavader]

There were some minor issues with finding the best local minimum in the smoothed score distribution, which is used for dynamic thresholding.

• Previously, local minima and maxima were only recognised when there was an immediate flip in the slope sign between bins. Therefore, any areas with slope zero were disregarded. Now, these are recognised, and the average value within the plateau is saved as local minimum / maximum.
• The threshold used to be defined by the most extreme local minimum within the range of accepted threshold values. This was updated to be the most extreme local minimum within the two most extreme local maxima (in order to avoid picking a value that does not separate the two major peaks, which represent low and high confidence gene calls respectively).
• Added an assertion that min threshold > max threshold.