[SYCL] Add non-standard vec aliases back to fix SYCL-CTS

[AlexeySachkov]

SYCL-CTS are still using some aliases which were removed from SYCL 2020 specification, see KhronosGroup/SYCL-CTS#446. Those tests are part of our post-commit and to make it green, this commit temporary adds those aliases back.

Note that it is not entierly clear if those aliases should have been removed in the first place, see KhronosGroup/SYCL-Docs#335.

[bader]

@cperkinsintel, @intel/llvm-reviewers-runtime, please, review ASAP. We need this patch to fix post-commit validation.

[aelovikov-intel]

IMO, it would be better to revert KhronosGroup/SYCL-CTS#440 until the spec issue is resolved. Until that, the CTS is technically in contradiction with the spec and hence is wrong.

I am only basing that on the links provided in this PR's description and might be missing some bigger picture.

[bader]

KhronosGroup/SYCL-CTS#440 aligns CTS with the spec, so there is no contradiction and reasons for reverting this PR. We need other patches removing data types intentionally removed from the spec and fixes in the spec and CTS to add unintentionally removed data types. All these changes should go before DPC++ changes to avoid post-commit CI failures.

As alternative option we can disable broken tests in our CI, but I don't think this is preferred way.

[aelovikov-intel]

I don't know where I got that link, please ignore my comment.

[aelovikov-intel]

Discussed with Alexey offline. I agree that it's better to revert the change causing the regression in post-commit until the CTS and/or spec issue is addressed.