WIP 'opaque' pointers, no atomics yet

[FractalFir]

A very, very rough first pass at removing typed pointers. Sadly, pointer issues tend to cause a lot of LLVM assertion triggers, and those are not fun to debug(which is why this relatively small change took so long).

Once I get something fully working, I will go over the PR and try to clean it up.

[nnethercote]

I looked through this but I don't know anything about this code so I don't have much to say.

Should the three commits be squashed together? (Do they make sense in isolation?)

[FractalFir]

I looked through this but I don't know anything about this code so I don't have much to say.

Should the three commits be squashed together? (Do they make sense in isolation?)

I will probably squash them once this is fully done. Still a bunch of refactor work left: I used bitcasts in a few places where LLVM 7 ptrcasts(now removed) would make more sense.

The argument against squashing is that each commit works in isolation, so if I (god forbid!) introduced a regression, this would make bisecting it easier.

I am fine either squashing or not.

[nnethercote]

If commits make sense in isolation, it's good to keep them separate. It's just that the third commit says "clippy fixes". Those fixes should be merged into the appropriate earlier commits so that each commit passes clippy.