Besst

- An order of magnitude more efficient path traversal algorithm. It reduces the number of paths tried based on an DP (-ish) algorithm, therefore we can explore more of the graph in less time. This should give better quality scaffolds compared to older versions.
- Betteer scoring of edges when library distribution is positively skwed (long tail) by fit to lognormal distribution.
- Added better logging to Statistics.txt and fixed minor logging bugs.
- Added an output file repeats_log.tsv on the following format: `contig_accession length coverage cov/mean_cov(exp number of placements) lib_mean placable`. This file serves as som information to future repeat placement in scaffolding.
- Added parameters `-z_min` (lower coverage cutoff), `-max_contig_overlap` how many bases in contig ends that are merged between two adjacent contigs in a created scaffold (default 200).

- Ability to detect and scaffold with PE-contamination
- More efficient (in practice) way to find paths with "good score" in contig graph.
- Speed-ups in pathfinding algorithm
- Better inference of gap sized when library is skewed (log-normal fitting - extending GapEst's assumptions of normal distribution)
- Better inference of library insert size distribution on fragmented assemblies (based on theory from http://www.biorxiv.org/content/early/2015/08/04/023929.abstract)