1) Group pathing: It would be efficient. It would require "special case" treatment of individuals, which is something more appropriate to a final commercial build-- when a game is alpha, you don't take shortcuts that will break possible developments. Optimization of pet pathing is important-- just pit your animals and watch your framecount to see what I mean-- but I can understand waiting to do this until you know EXACTLY how you want to treat animals, slaves, liasons, squadmates, pursuing mobs, etc. It makes a lot of sense to me to start out with logic-expensive routines and optimize as necessary. (Not that anybody really thinks this is a commercial game. It'll be under development forever, and that's a good thing.)
Skipping player-designated paths.
3) Automated path-mapping: Makes a lot of sense with a static fortress-- but these aren't static. There's the potential for the cost being greater than the benefit. Yeah, Copperblazes would do well with some automated path-mapping. But an initial fortress would take a hit. Not that I don't think it's a good idea. Initial fortresses have fps to spare; established fortresses do not.
2) Player-designated paths: Steeper learning curve. Not really a bad idea though; it preserves basic DF while allowing fps increase.
Mostly? It's boring junk to program. If this were a commercial game, sure, somebody would have to do this crap. But given that DF is some guy who's good at what he does and wants to have fun doing it, who can blame him for prioritizing procedurally generated cybermagic gas-exchange organs? And really, would you rather have a few more fps instead of that?