avx512_target_feature is now stable on nightly

[folkertdev]

and enabling it causes errors on CI

[rustbot]

r? @Amanieu

rustbot has assigned @Amanieu.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[folkertdev]

Just writing up what we found here (yet more reasons to make this a submodule): rust-lang/rust#127013 changed the formatting of f16 values (to something much more helpful) in rust. In C, and f16 is apparently formatted as a hexadecimal value. So we see results like this in the intrinsics tests:

C: result -1: float16x4x2_t(float16x4_t(0x0000, 0x0000, 0x37ff, 0x37ff), float16x4_t(0x0400, 0x0400, 0x3800, 0x3800))
Rust: result -1: float16x4x2_t(float16x4_t(0.0, 0.0, 0.4998, 0.4998), float16x4_t(6.104e-5, 6.104e-5, 0.5, 0.5))

I think the quick fix is to make stdarch have the old formatting behavior, so I implemented some hacks to make that happen. Long-term, hopefully there is a better way to compare the C and rust output?

[Amanieu]

f16 vector should use the same default formatting as f16. Could you change this to only apply to the Rust output of intrinsic-test instead of all users of the Debug impl?