[clang] Fix Strong VTables Accidentally Emitted by #159856 (PR #200942)

[Eli Friedman via cfe-commits]

efriedma-quic wrote:

> before it's supposed to be called

I'm not sure what this means; are you referring to the way it's forcing the vtable to be emitted?

> because changes to that code would affect code generation for all targets, not just OpenMP offload.

I expect they'll eventually end up entangled anyway.  Trying to layer logic like this almost always leads to more bugs, and the logic just ends up harder to understand because it's scattered across multiple different areas.

https://github.com/llvm/llvm-project/pull/200942