Rollup merge of rust-lang#60191 - gnzlbg:f16c, r=alexcrichton

Add f16c target_feature

This is requires for Intel 16-bit half-precision float intrinsics: https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=fp16&expand=1769 - see companion stdsimd PR: rust-lang/stdarch#737.

LLVM, Wikipedia CPUID page, and the Intel Dev Manual all call this CPUID feature "F16C", but the Intel intrinsics guide calls this "FP16C" - this is probably a bug in the intrinsics guide which we are tracking here: rust-lang/stdarch#738

r? @alexcrichton

src/librustc_codegen_llvm/llvm_util.rs

@@ -145,6 +145,7 @@ const X86_WHITELIST: &[(&str, Option<&str>)] = &[
     ("bmi1", None),
     ("bmi2", None),
     ("cmpxchg16b", Some("cmpxchg16b_target_feature")),
+    ("f16c", Some("f16c_target_feature")),
     ("fma", None),
     ("fxsr", None),
     ("lzcnt", None),