BUG: Fix build issues on SPR and avx512_qsort float16

[r-devulap]

Fixes #25274, #25382

1. Marks explicit template specializations in x86-simd-sort as inline.
2. Adds CI to run tests on TGL and SPR platforms with latest version of Intel SDE v9.27.
3. Bug fix in avx512_qsort<float16>, which was detected when enabling the CI on TGL.

[r-devulap]

Enabling tests on SDE found a bug in avx512_qsort, which this patch fixes.

[seiko2plus]

It should also resolve #25382, Isn't?

[r-devulap]

It should also resolve #25382, Isn't?

Yup, I found it when enabling the CI on Intel SDE TGL platform. It will be useful to run tests on TGL and SPR.

[r-devulap]

intel_sde CI takes about 35 min to complete, let me know if you want me to split it. We could run TGL and SPR tests in separate CI jobs.

[seiko2plus]

intel_sde CI takes about 35 min to complete, let me know if you want me to split it. We could run TGL and SPR tests in separate CI jobs.

Yes, it would be better also if the baseline of this test is set to avx512_spr so we can test static dispatching.

[r-devulap]

intel_sde is split into intel_sde_avx512 (cpu-baseline avx512f) and intel_sde_spr (cpu-baseline avx512_spr)

[r-devulap]

Something went wrong with the Qemu tests. Will close and open to re-run them.

[seiko2plus]

Thank you Raghuveer.

[charris]

Bug fix in avx512_qsort, which was detected when enabling the CI on TGL.

@r-devulap Is this bug present in 1.26.x? If so, could you make a bug fix PR for it? I don't want to try taking this whole PR back, the files have been renamed and reorganized.

[r-devulap]

@charris see #25475.