Update more math intrinsics to use LLVM instead of libm

[ZuseZ4]

As a prototype to be discussed on zulip. If accepted I'd update miri, f16/f128, clean it up, and add the other ones tan, asin, acos, atan, sinh, cosh.

Motivation: For HPC/Scientific Computing Benchmarks I often port existing C/C++ code to Rust and compare how they fare. When looking at IR differences, these often came up. C++ has the -fno-math-errno flag, but we explicitly don't guarantee which implementation we use, so I think we are free (and should) default to this behaviour which allows more optimizations.

[rust-log-analyzer]

The job tidy failed! Check out the build log: (web) (plain enhanced) (plain)

Click to see the possible cause of the failure (guessed by this bot)

fmt: checked 7209 files
Diff in /checkout/library/std/src/num/f64.rs:1066:
     #[stable(feature = "rust1", since = "1.0.0")]
     #[inline]
     pub fn tanh(self) -> f64 {
-       intrinsics::tanhf64(self)
+        intrinsics::tanhf64(self)
     }
 
     /// Inverse hyperbolic sine function.
Bootstrap failed while executing `test src/tools/tidy tidyselftest --extra-checks=py,cpp,js,spellcheck`
Currently active steps:

[tgross35]

Zulip discussion starting at #general > algebraic float ops @ 💬

[rust-bors]

☔ The latest upstream changes (presumably #162148) made this pull request unmergeable. Please resolve the merge conflicts by rebasing.