Remove HasCompact::Type bounds

[ascjones]

Removes the requirement for HasCompact::Type: TypeInfo bounds I had to add to some impls in substrate, see them removed in this commit paritytech/substrate@7da3bb8.

It does this by just registering T for a compact field in a Compact<T>. This relies on the fact that the blanket impl always has Type = Compact<T>.

@dvdplm maybe you already tried this as part of #53 and decided against it.

[dvdplm]

Yeah so this is what I had in my very first implementation. I was wary about it mostly because I had a poor understanding of what Compact and HasCompact even was; it felt a bit too easy to be correct?

When a type is #[code(compact)] it means that instead of encoding the type we actually encode <the_type as HasCompact>::Type so it becomes a bit philosophical here: do we want TypeInfo for the_type or for the type actually used in the encoding, ::Type?
At the time of #53 I felt pretty convinced that we wanted TypeInfo for the type that would end up being encoded (and then we ran into fun compiler bugs and I didn't think too deeply about it).

If this works for substrate then I'm all for it.

[ascjones]

do we want TypeInfo for the_type or for the type actually used in the encoding, ::Type

TypeInfo for ::Type, however we know this is always Compact<T> because of the blanket impl.

I'll leave this here as a draft for now, and come back to it once the substrate PR is in a good state, and maybe explore other ways to get around having those bounds. And consider whether this is philosophically/practically the right approach.

[dvdplm]

LGTM