Add support for (inlined) destructors

[PathogenDavid]

There isn't a way I'm aware of to convince MSVC or Clang to export an inline destructor without modifying its definition.

Destructors were disabled from module definition generation entirely (even non-inline ones) in the prototype and it's unclear exactly why that is.

Worth noting: Virtual destructors are always available in the output.

[PathogenDavid]

Found out why destructors were disabled. clang_Cursor_getMangling returns the wrong mangling for destructors on Windows. You have to use clang_Cursor_getCXXManglings instead. In theory constructors have the same problem, but Microsoft's mangling doesn't differ between different types of constructors.

---

Details from my notes:

This originally came up because Clang is returning the wrong mangling for the ImFont destructor. Clang returns ??_DImFont@@QEAAXXZ, but the exported symbol is ??1ImFont@@QEAA@XZ.

Looking at mangleCXXDtorType, the difference between the two is that ?_D is for a vbase destructor and ?1 is for a normal destructor.

We use clang_Cursor_getMangling to get the mangling. This function is effectively implemented by ASTNameGenerator::Implementation::writeName, which for function declarations is implemented by writeFuncOrVarName which for a destructor always uses the Dtor_Complete destructor type, which will be the ?_D mangling.

Looking at libclang, there's a separate clang_Cursor_getCXXManglings which returns multiple strings. This function effectively calls ASTNameGenerator::Implementation::getAllManglings. With a Microsoft ABI this will only return a single value for destructors, Dtor_Base (which is the ?1 version we expected.)

[PathogenDavid]

Between 756b969 and 23c4edd, I am calling this done.

Better support for deletion is tracked by #159

Improving friendliness of constructing/destructing C++ objects is tracked by #161