[llvm] [LangRef] make consequences of NaN rules for pow(i) more explicit (PR #170177)

[Eli Friedman via llvm-commits]

================
@@ -16831,6 +16836,11 @@ trapping or setting ``errno``.
 When specified with the fast-math-flag 'afn', the result may be approximated
 using a less accurate calculation.
 
+Note that due to how :ref:`LLVM treats NaN values <floatnan>`, the special cases
----------------
efriedma-quic wrote:

> I think we should be very explicit about this being non-determinism

Sure.  Maybe "note this operation can non-deterministically return a non-NaN result given an SNaN input"? Not really attached to the exact wording.

https://github.com/llvm/llvm-project/pull/170177