revert post-zoonomia outgroup changes

[glennhickey]

og-selection is impactful high up in deep, large trees. the 10-way test i've been using doesn't really cover it, nor do recent projects like primates or apes.

even with some new machinery (#1875), any errors in the tree can wildly swing outgroups between clades. going back to:

• allowing ancestral outgroups should mitigate this, as each outgroup can potentially cover far more of the tree)
• and using 3 outgroups further reduces the odds of missing anything important.

(pending a few focused tests to merge this or #1875)