Expose various new Create and conversion APIs for the Vector types

[tannergooding]

This resolves #92299 and #94384

[ghost]

Note regarding the new-api-needs-documentation label:

This serves as a reminder for when your PR is modifying a ref *.cs file and adding/modifying public APIs, please make sure the API implementation in the src *.cs file is documented with triple slash comments, so the PR reviewers can sign off that change.

[ghost]

Note regarding the new-api-needs-documentation label:

This serves as a reminder for when your PR is modifying a ref *.cs file and adding/modifying public APIs, please make sure the API implementation in the src *.cs file is documented with triple slash comments, so the PR reviewers can sign off that change.

[tannergooding]

With this PR, the only remaining simdashwintrinsic importation handling is for Vector<T>.

After this PR is merged, we can take advantage of that fact and change the importation logic a bit. Namely, since we've moved towards having parity across Vector<T> and Vector64/128/256/512<T> we can special case Vector<T> to import as if it were Vector128/256/512<T> instead (based on its size), allowing us to remove simdashwintrinsic in its entirety and just defer down a single path.

[martenf]

Will
public static Vector512<T> Create<T>(Vector128<T> value) => Create(Vector256.Create(value, value));
and
public static Vector512<T> Create<T>(Vector256<T> value) => Create(value, value);

use _mm512_set4_ for the appropriate types on AVX512 systems or classical broadcasts?

[tannergooding]

use mm512_set4 for the appropriate types on AVX512 systems or classical broadcasts?

Not in this PR, but I will be getting it updated in a separate PR. There is some additional work required to get the different overloads of Create to be imported and optimized correctly; so this PR mostly focuses on exposing the newly approved APIs without regressing the current codegen.

[tannergooding]

CC. @dotnet/jit-contrib, this is ready for review. Removes the rest of the Vector2/3 handling from simdashwintrinsic and adds some of the new xplat APIs that were approved.

[EgorBo]

Nice! Glad that Vector<T> now has Create

Also, e.g. in #86220 @xoofx had to do AsVector128().AsVector3() for V4 to V3 conversion, now there is AsVector3 for it.

[LoopedBard3]

Non-mono additional regressions:
Linux x64: dotnet/perf-autofiling-issues#36405
Linux Arm64: dotnet/perf-autofiling-issues#36655, dotnet/perf-autofiling-issues#36654
Windows Arm64: dotnet/perf-autofiling-issues#36661

[tannergooding]

The "regression" here is because we aren't constant folding the call to SinCos. It's an improvement for the real world scenario

[lewing]

mono wasm interpreter regressions here dotnet/perf-autofiling-issues#36484

cc @kg @BrzVlad