Implement *actual* polyhedral loop transformation

Just got invited to a Dagstuhl seminar put on by the actual polyhedral crowd. I think it might be a nice nod to them to make it so loopy can do something that fits with the classical polyhedral model. Given that we're on top of ISL already, I don't even imagine that that'll be super hard. Plus I've always been curious what that's all about.

cc @mattwala (Not that I'm suggesting that as an additional thing to tackle--just FYI.)

cc @isuruf (Any interest? Any clue what this is all about? Wikipedia)