[clang-doc] Reconsider use of enum InfoType::IT_default

[evelez7]

The use of InfoType::IT_default seems to clash with LLVM style which prefers to warn if not all enums are covered in a switch without default. We use our own default enum, which a.) could just be covered by default and b.) might prevent us from seeing the warning.

It doesn't look like we set it anywhere besides when default constructing an Info, so I doubt that it's actually checked before a proper type is set. It's not checked in the unit tests where it is explicitly passed to construct a reference.

[ilovepi]

Certainly we should prefer to use fully covered switches where possible, so that we get compile time errors/warnings when the enum is updated. My memory is that those diagnostics only work for scoped enums (enum class), though... do we need to update our definitions?

[evelez7]

The warning warns on enum and enum class from my testing with clangd.

[evelez7]

Another issue I find with the IT_default and InfoType is that it's semantically confusing. Does InfoType denote the constructs we document in source or the Info data structures we have implemented?

For example, struct TypeInfo doesn't hold its own enum class InfoType. Its held by its member Reference RefType. But return types (which are TypeInfo) are classified with IT_default while record references (e.g. child records) get IT_record. If we are documenting a type, like a return type, then maybe there should be a IT_type.

There should be a mega-nit tag :)

[evelez7]

Fixed with #143919.