Fixed issue with FAR and FRR values being exactly 1 or 0, respectively

I found a small bug when FAR=1 was used in far_threshold, and FRR=0 in frr_threshold. This resulted in a DIR plot that had a random last value (as it accessed memory at an uninitialized location).

This fixes it. There should be no implication for any other plot done so far.

mentioned in commit 5c508b9a

merged

Thank you for the quick fix, @mguenther