[libc][math][c23] Improve C23 math function rsqrtf16

[amemov]

Continuation of #132818 . In #137545, an implementation was introduced that calls hardware sqrt() to compute rsqrtf16 - which is turned out to be the fastest option ( approximations were slower by ~30% ). But, there is still some work left - for targets that don't have fixed-precision floating points, an int-based approximation is still needed.

I would like to work on this, please assign me to it. Thank you!

[llvmbot]

@llvm/issue-subscribers-libc

Author: Anton Shepelev (amemov)

Continuation of #132818 . In #137545, an implementation was introduced that calls hardware sqrt() to compute rsqrtf16 - which is turned out to be the fastest option ( approximations were slower by ~30% ). But, there is still some work left - for targets that don't have fixed-precision floating points, an int-based approximation is still needed.

I would like to work on this, please assign me to it. Thank you!