Add unified set of Diffusion model transforms

[Beinsezii]

Currently just for sampling.functional

Once I add velocity this can trivially go into the diffusers wrapper instead of a raw prediction function, and then moved through structured

Possibly I could associate the schedule and therefore also the sigma transform type base_schedule = Linear or whatever. Will have to play around see if FlowModel can be made to handle a Variance-Preserving schedule and vice-versa.

Just so fucking glad I don't have to switch on sigma_transform anymore for the tableau solver.

[Beinsezii]

so it seems to be impossible to represent velocity with this method without performing at least one output conversion, which means the only "public" parameters we can have is forward, backward, to_x, and from_x. Eta, Z, H, can't be public because they may misalign with the conversion.

I guess in theory that "could" work, and you could argue that velocity is an invalid intermediate because it looks like garbage when used in higher order VP but in my mind if we aren't going to realistically be able to use DiffusionFlow's prime algorithm because the terms can't be guaranteed then we may as well make a different algorithm that doesn't have these limits #57 where even if the terms are much less useful on their own, since they can be guaranteed on all models there it's better than not having them at all IMO.