compiler_rt: enable fmax and fmin to distinguish between 0.0 and -0.0

[papparapa]

Closes #24214

This PR enables fmax and fmin to differentiate between 0.0 and -0.0, making it compatible with musl.

Maybe we can remove the musl fmin and fmax that are not processor specific and that would be another PR.

[rpkak]

Info from testing with zig-libc-test (on x86_64-linux-musl):

The test-cases fmax, fmaxf, fmaxl, fmin, fminf and fminl pass with zig master.

Without these lines:

zig/src/libs/musl.zig

Lines 924 to 929 in 4f40490

	"musl/src/math/fmax.c",
	"musl/src/math/fmaxf.c",
	"musl/src/math/fmaxl.c",
	"musl/src/math/fmin.c",
	"musl/src/math/fminf.c",
	"musl/src/math/fminl.c",

they fail with zig master, but pass with this PR.