[clang] [compiler-rt] [llvm] [CMake][compiler-rt] Support for using compiler-rt atomic library (PR #106603)

[Chris Apple via cfe-commits]

cjappl wrote:

> I'm seeing a test failure but I'm not sure if it's an issue with RTSan or compiler-rt atomic implementation:

Interesting, so that test basically tests that large atomics that insert locks under the hood die as expected. Zooming in on the second part of the test:


```
std::atomic<std::array<float, 2048>> large_atomic;
  ASSERT_FALSE(large_atomic.is_lock_free());     ///// HERE IS INTERESTING!
  auto Func = [&]() {
    std::array<float, 2048> x = large_atomic.load();
    return x;
  };
  ExpectRealtimeDeath(Func);
  ExpectNonRealtimeSurvival(Func);

```

It seems like the compiler-rt version says it is lock free, but then we aren't dying. Does it take another lock, different from a pthread_mutex_lock? Those are the locks we look for in our implementation.

https://github.com/llvm/llvm-project/pull/106603