Consider reporting an ambiguity if both modules and primitives have the same name in an intra-doc link

[jyn514]

Originally this defaulted to the module. #75318 changed this to default to primitives since seemed to be the less confusing behavior (#58699). However I'm wondering if the real issue is that rustdoc didn't report an ambiguity like it normally would for multiple namespaces. I think it would be better to say 'both module@char and type@char are in scope, pick one.

[jyn514]

Note the ambiguity here is because both are in the type namespace. The way rustdoc currently finds ambiguity is if there are multiple namespaces with a match, but because these are in the same namespace, it doesn't notice.

[jyn514]

This needs to happen before stabilization (#74430) since it's not backwards compatible.

[jyn514]

Ugh, this breaks type@char because both modules and primitives are in the type namespace. Maybe we could introduce a primitive@ or prim@ disambiguator? There was some talk of that for #63351. That seems a little silly to do just for this though ...

[jyn514]

The other alternative is to special case type@char as meaning the primitive, which is what happens currently. But that seems pretty hacky.

[jyn514]

Going to introduce prim@char for now and we can discuss it more during review.