Optimize AsVector, AsVector128, GetUpper, As and WithElement with ARM64 intrinsics

[kunalspathak]

Optimize some more APIs with ARM64 hardware intrinsics:

• Optimize Vector128.AsVector ()
  AsVector_before.txt
  AsVector_after.txt

• Optimize Vector.AsVector128()
  Vector_AsVector128_before.txt
  Vector_AsVector128_after.txt

• Optimize Vector128.GetUpper() in JIT
  getupper_before.txt
  getupper_after.txt

• Optimize Vector64.As
  As_before.txt
  As_after.txt

• Optimize Vector64.WithElement()
  withelement_vector64_before.txt
  withelement_vector64_after.txt

• Optimize Vector128.WithElement()
  withelement_vector128_before.txt
  withelement_vector128_after.txt

Contributes to #33308 and #33496.

[kunalspathak]

Thanks @tannergooding for reviewing.

@echesakovMSFT , @TamarChristinaArm.

[TamarChristinaArm]

Thanks @kunalspathak! I think you have a ticket for this right?

        53003C00          uxth    w0, w0
        4E0A1C10          ins     v16.h[2], w0

That there's no zero extension needed here? I think I remember this being discussed when Insert was added?

[kunalspathak]

Thanks @kunalspathak! I think you have a ticket for this right?

        53003C00          uxth    w0, w0
        4E0A1C10          ins     v16.h[2], w0

That there's no zero extension needed here? I think I remember this being discussed when Insert was added?

Yep, its here. #35590 (comment)