wflign for long global sequence alignment

[ekg]

This update changes the alignment model in edyeet to wflign. At a low level, edlib is still used to derive base-level alignments.

In result, the alignment of long sequences of many megabases is now possible in low memory and time.

wflign is tolerant of relatively high divergence over the full sequence length, which results in a higher fraction of alignments. For a given setting of -p and -a, we derive more final alignments at the chosen -p threshold.

The meaning of -a[%], --align-pct-id=[%] changes. Now, it is a simple filter on the output alignments. The -p[%], --map-pct-id=[%] parameter and new wflign reduction parameters structure the alignment. It continues to be set by default to be equal to -p, which may be overly restrictive.

todo:

• [ ] It may be important to update wflign to not run WFA when the aligned segments are shorter than some small multiple of the segment length.
• [ ] When we are given targets that aren't correct, we will spend a very long time aligning. Computation of wflign should have an upper limit of difficulty appropriate for the chosen map-pct-id.

[subwaystation]

oh yeah :fire: looking forward to see this in pggb :)

[ekg]

There is an other change that will be noted. Alignments are now emitted in small chunks. This is a feature of the wflign algorithm.

[ekg]

And, we now build with cmake (see README).