Skipped ahead a few pages, but, phew, this is a busy thread.
There's a very good reason for not leaning too hard on what was done before. Mostly at least in 1e and 2e, because what was done before was a horrible, broken mess that was one of the worst sets of mechanics released for either edition. 2e psionics were so broken that you actively had to try not to break the mechanics. They were just flat out awful.
In fact, they were so bad, that I completely skipped psionics in 3e. I have no idea what psionics look like in 3e and I don't care. If psionics requires an entirely new set of mechanics walled behind an entire book, then, well, I hope psionics never comes to 5e. It was garbage back then, and it will be garbage now.
As a DM, no, I'm not going to read an entire book just for your character. Sorry. Not interested. So, any approach to psionics that creates a complete class and whatnot is a total non-starter for me. I will treat it exactly the way I treated psionics in 3e, 3.5e and 4e - it doesn't exist as far as I'm concerned.
And, the same goes for Dark Sun. If I need to buy an entire psionic book just to run Dark Sun, well, again, I'll treat it exactly the way I did in the 90's. Completely ignore it.
So, for me, the sub-class approach is the best way. We get psionics. We don't have to bolt on completely different mechanics, just so one player at the table can get to feel all special. If you can't compromise and meet half way, then, well, there's a ton of 3rd party material out there for you to use.
Good grief, Warlord fans get pooped on for wanting a class that would take up all of three or four pages and would use existing mechanics. We get told that it's too hard, doesn't fit in the game and we should be satisified with what we've already got. Meanwhile psionic fans are demanding an entire book?!?! And they think this is a reasonable goal?
Good luck with that.