[llvm] specify NaN behavior more precisely (PR #66579)

[Joshua Cranmer via llvm-commits]

https://github.com/jcranmer-intel commented:

It is worth calling out somewhere that qNaN/sNaN are defined using the convention that leading bit = 0 is sNaN, 1 is qNaN, and systems that do not adhere to this convention (e.g., old MIPS) are unsupported.

Also probably worth calling out that things like `bitcast` do not count as floating-point operations.

https://github.com/llvm/llvm-project/pull/66579