Do not generate bindings if replacement type has a different size/alignment than replacee

[fitzgen]

This would generate really hard to debug issues.

[emilio]

This is non-trivial, but I'm not sure it's the path forward. Indeed, a good use case for replacements is when bindgen can't know/infer enough layout information for the replacee. A patch solving this would pretty much break that use case.

I think we should spit an error! instead, what do you think?

[fitzgen]

Argh. Chicken and egg kind of problem, isn't it?

Maybe only fail if we do understand the replacee's layout and we know the replacement's layout is different?

[fitzgen]

It's worth noting that we tend to be better about turning things we don't understand into TypeKind::Opaque since this issue was filed, and so this and #754 are probably more feasible now than #188 (comment) makes it seem.