Add Needleman-Wunsch alignment

[timholy]

This is a special case in which no gaps are allowed in one of the
"sequences." The intent is to use this for aligning the TM helices of
GPCRs: the "reference sequence" is just the 14 residues at the start and
end of the reference structure's TM helices, and the "query sequence" is
the full sequence of another 7TM.

The core algorithm is supplied with a pairwise cost matrix. Convenience
methods allow you to compute the cost from sequence. Currently, cost can
be computed either from distance alone or distance + orientation of the
residues.

[tmcgrath325]

This looks good to me! Based on a handful of receptors, it looks like the constrained NW alignment to rhodopsin does a good job of identifying TMs

[timholy]

Great! I manually stepped through all 230 V1Rs, and identified 15 that seem to have issues. That seems pretty good, and we can exclude those 15 from further analysis. That might be a separate PR to a different repo.