Change f32::midpoint to upcast to f64

This has been verified by kani as a correct optimization

see: #110840 (comment)

The new implementation is branchless, and only differs in which NaN
values are produced (if any are produced at all). Which is fine to change.
Aside from NaN handling, this implementation produces bitwise identical
results to the original implementation.

Author: RustyYato

library/core/src/num/f32.rs

@@ -1016,26 +1016,7 @@ impl f32 {
     /// ```
     #[unstable(feature = "num_midpoint", issue = "110840")]
     pub fn midpoint(self, other: f32) -> f32 {
-        const LO: f32 = f32::MIN_POSITIVE * 2.;
-        const HI: f32 = f32::MAX / 2.;
-
-        let (a, b) = (self, other);
-        let abs_a = a.abs_private();
-        let abs_b = b.abs_private();
-
-        if abs_a <= HI && abs_b <= HI {
-            // Overflow is impossible
-            (a + b) / 2.
-        } else if abs_a < LO {
-            // Not safe to halve a
-            a + (b / 2.)
-        } else if abs_b < LO {
-            // Not safe to halve b
-            (a / 2.) + b
-        } else {
-            // Not safe to halve a and b
-            (a / 2.) + (b / 2.)
-        }
+        ((f64::from(self) + f64::from(other)) / 2.0) as f32
     }
 
     /// Rounds toward zero and converts to any primitive integer type,