P2615 (Meaningful exports) is not fully implemented: Diagnostic required

Nobody has claimed this yet.

Test code ([CE](https://godbolt.org/z/Yov6xcKq7)):

```cpp
export module M;
export template struct S {};

// ❌ - Clang/MSVC/EDG: OK
// ✅ - GCC:
// - error: declaration of partial specialization in unbraced
// export-declaration
// - note: a specialization is always exported alongside its
// primary template
export template struct S {};

// ❌ - Clang/MSVC/EDG: OK
// ✅ - GCC: error: explicit instantiations are not permitted here
export template struct S;

// ❌ - Clang/MSVC/EDG: OK
// ✅ - GCC:
// - error: explicit specializations are not permitted here
// - note: a specialization is always exported alongside its
// primary template
export template <> struct S {};
```

[P2615R1](https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2022/p2615r1.html) forbids applying `export` directly to partial specializations and explicit instantiations/specializations, as doing so is meaningless. (These declarations are allowed to appear within an `export` block for convenience.) The code above should be diagnosed accordingly.

Related: #147135.

Start with the Compiler Explorer test code in the issue and read P2615R1, focusing on the rules for export declarations involving partial specializations, explicit instantiations, and explicit specializations. Compare the current diagnostics with the three examples; done means each directly exported declaration is diagnosed as required, while declarations inside an export block remain permitted.

Written by the indexing model from the issue text.