Also adds the score dropoff to PE path

[ksahlin]

PR #601 fixed only for the SE path, plus for rescue function in PE mode. This PR completes the changes to also use score-based dropoff for the pairing of NAMs(/chains).

Neither runtime nor accuracy changes significantly (see accuracy experiment below). The fix is for consistency in approach.

Paired-end alignment, 200k PE reads

Genome	Len	Tool	Acc%	Δ vs main
chrY	100	main	44.259%
chrY	100	score-dropoff	44.279%	+0.020%
chrY	250	main	55.867%
chrY	250	score-dropoff	55.870%	+0.004%
chrY	500	main	67.846%
chrY	500	score-dropoff	67.842%	−0.004%
CHM13	100	main	93.525%
CHM13	100	score-dropoff	93.527%	+0.002%
CHM13	250	main	95.396%
CHM13	250	score-dropoff	95.396%	+0.000%
CHM13	500	main	96.919%
CHM13	500	score-dropoff	96.919%	+0.000%
Maize	100	main	87.396%
Maize	100	score-dropoff	87.378%	−0.018%
Maize	250	main	95.942%
Maize	250	score-dropoff	95.943%	+0.001%
Maize	500	main	98.442%
Maize	500	score-dropoff	98.444%	+0.002%

[marcelm]

I confirm that there is nearly no difference in accuracy, and no discernible difference in runtime.