Speed up how node names are processed on initialization

[fedarko]

Just making a record of this here so we don't forget it. (@kwcantrell and @ElDeveloper did all the work of finding this!!!)

Relates to the tree shown in this twitter thread. Basically, when there are a ton of node names, this block of code:

empress/empress/support_files/js/empress.js

Lines 463 to 472 in 1b7a8d3

	var nodeNames = this._tree.getAllNames();
	// Don't include nodes with the name null (i.e. nodes without a
	// specified name in the Newick file) in the auto-complete.
	nodeNames = nodeNames.filter((n) => n !== null);
	// Sort node names case insensitively
	nodeNames.sort(function (a, b) {
	return a.localeCompare(b, "en", { sensitivity: "base" });
	});
	nodeNames = _.uniq(nodeNames);

... slows things down a lot, because string comparisons are slow.

There are many possible ways we could handle this (better algorithms / data structures in JS, preprocessing stuff in Python, etc.) -- it's probably most important right now just to implement something that fixes this, since this is a pretty significant bottleneck.