seqwish implements a lossless conversion from pairwise alignments between
sequences to a variation graph encoding the sequences and their alignments.
As input we typically take all-versus-all alignments, but the exact structure
of the alignment set may be defined in an application specific way. This
algorithm uses a series of disk-backed sorts and passes over the alignment and
sequence inputs to allow the graph to be constructed from very large inputs
that are commonly encountered when working with large numbers of noisy input
sequences. Memory usage during construction and traversal is limited by the use
of sorted disk-backed arrays and succinct rank/select dictionaries to record a
queryable version of the graph.