[release/7.0-rc1] Vectorize {Last}IndexOf{Any}{Except} without code duplication

[adamsitnik]

Backport of #73768 to release/7.0-rc1

I'm backporting this as a special case (I missed the cutoff by hours), as discussed with @stephentoub, who discussed it with @danmoseley.

This PR gives us:

• great validation of the new Vector128 and Vector256 APIs.
• vectorization of all Span.{Last}IndexOf{Any}{Except} methods which translates to impressive performance improvements (the larger the input and more work needed, the better).

The performance has been verified on arm64 Windows and Linux, x64 Intel and AMD. Full results can be found here

[ghost]

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

Issue Details

---

Backport of #73768 to release/7.0-rc1

I'm backporting this as a special case (I missed the cutoff by hours), as discussed with @stephentoub, who discussed it with @danmoseley.

This PR gives us:

• great validation of the new Vector128 and Vector256 APIs.
• vectorization of all Span.{Last}IndexOf{Any}{Except} methods which translates to impressive performance improvements (the larger the input and more work needed, the better).

The performance has been verified on arm64 Windows and Linux, x64 Intel and AMD. Full results can be found here

Author:	adamsitnik
Assignees:	-
Labels:	area-System.Memory, tenet-performance
Milestone:	7.0.0

[danmoseley]

I believe this arguably falls under "1. Broken or significantly incomplete 7.0 scenario" as vectorization of API is a significant focus for us this release, and this is part of work @adamsitnik is doing to ensure our approaches align with best practices guidance he is drafting.

So I'm comfortable taking this this week, but not after this week.

[carlossanlop]

Approved and signed-off.
CI passed.
Ready to merge.