Vectorize TensorPrimitive generic operations for sizeof(T) < 4

[stephentoub]

#96451 added generic versions of the existing float-based operations, complete with vectorization. However, that vectorization is currently skipped if sizeof(T) < 4, which means it doesn't kick in for byte, sbyte, ushort, short, or char. That should be implemented (or we should explicitly decide we don't need to).

runtime/src/libraries/System.Numerics.Tensors/src/System/Numerics/Tensors/netcore/TensorPrimitives.netcore.cs

Line 15 in 7e51126

// - Vectorize integer operations when sizeof(T) == 1 or 2 (currently only vectorized in most operations for sizeof(T) == 4 or 8).

Tagging subscribers to this area: @dotnet/area-system-numerics-tensors
See info in area-owners.md if you want to be subscribed.

Issue Details

---

#96451 added generic versions of the existing float-based operations, complete with vectorization. However, that vectorization is currently skipped if sizeof(T) < 4, which means it doesn't kick in for byte, sbyte, ushort, short, or char. That should be implemented (or we should explicitly decide we don't need to).

runtime/src/libraries/System.Numerics.Tensors/src/System/Numerics/Tensors/netcore/TensorPrimitives.netcore.cs

Line 15 in 7e51126

// - Vectorize integer operations when sizeof(T) == 1 or 2 (currently only vectorized in most operations for sizeof(T) == 4 or 8).

Author:	stephentoub
Assignees:	-
Labels:	area-System.Numerics.Tensors, tenet-performance
Milestone:	9.0.0