Caterpillar

Type inference of recursive functions is causing me a lot of trouble. And I haven't even mentioned mutual recursion yet! That complicates things further, of course.

After I gave up on processing branches based on a queue (for now; maybe I'll give it another go later), I've been looking into building a call graph of branches involved in recursion, and figuring out the right order of processing based on that. I already tried it earlier, before looking into the queue-based approach. Now I've come back to it, with some new ideas.

But first, it's compiler cleanup time again! So far, anonymous functions are being handled in a very sub-optimal way. I didn't notice, because it hasn't been causing any major problems. But with this new challenge, the issue has become obvious. So yeah, that's were I'm at right now. I'm sure it'll work out one way or another, and a cleaner compiler pipeline is always a nice bonus.