[llvm] [FMV][GlobalOpt] Bypass the IFunc Resolver of MultiVersioned functions. (PR #87939)

[Andrew Carlotti via llvm-commits]

andrewcarlotti wrote:

The testcase and description look valid, although the description seems a little more conservative than necessary.

If you were to add a `bar._Mmops` version, would it's call be optimised to `test_multi_bb_resolver._Mmops`? And would that also enable the call from `bar.default` to be optimised to point to `test_multi_bb_resolver.default`?

https://github.com/llvm/llvm-project/pull/87939