[llvm] ValueTracking: Identify implied fp classes by general fcmp (PR #66505)

[Adrian Kuegel via llvm-commits]

akuegel wrote:

> @akuegel remember this test failure in TF last week? maybe you have additional information here?

I have debugged now which class is detected, and it is `FPClassTest::fcPosInf`. Does that help?
I see that you added this code before relanding:

```
if (RHSClass == fcPosZero || RHSClass == fcNegZero)
    return {nullptr, fcAllFlags, fcAllFlags};
```

Any reason why you don't just treat fcPosInf, fcNegInf, etc. the same way? It seems the original assert was written with some assumption that doesn't hold.

https://github.com/llvm/llvm-project/pull/66505