new options for matUtils extract: --max-mut-density, --node-stats

[AngieHinrichs]

I added two options to matUtils extract to prune problematic sequences and branches from the SARS-CoV-2 tree before making a minimized tree for use with pangolin:

• --node-stats filename writes out various statistics of each internal node and sample, including the number of reversions to reference since the most recent ancestor annotated as a clade/lineage root. (For recent pangolin data releases, I've been pruning samples with 2 or more reversions since lineage root.)
• --max-mutation-density D removes samples descended from nodes whose "mutation density" (sum of mutation counts divided by number of leaves) is greater than D (I've been using D=2 for pangolin), with several exceptions:
  • don't mark an annotated clade root for removal even if it or an ancestor has a high mutation density
  • keep samples descended from a node with mutation density < 1/D even if an ancestor of that clade is marked for removal, so tight clusters are kept even if on an otherwise poor-quality branch
  • keep all samples from annotated clades with fewer than 150 leaves (may want to add an option to set that someday)
  • divide an internal node's own mutation count by its leaf count when summing with descendants' mutation counts, so we don't over-penalize one long branch to a tight cluster.

[yatisht]

@jmcbroome I am going to merge this, but since --node-stats filename only writes summary stats, maybe it's better suited to matUtils summary? I'll let you decide if you want to move it there or not.

[jmcbroome]

re: Angie- thanks for getting this in! I'll check it out and see about using it with the automated lineage output.

re: Yatish- I agree that matUtils extract is extremely bloated already. I'll look into reorganizing- thanks