MemoryMarshal.IndexOfAny/LastIndexOfAny for non-primitives does not special-case small needle spans

[GrabYourPitchforks]

Based on a conversation at #53652 (comment).

In the implementation of MemoryExtensions.IndexOfAny<T>(ROS<T>, ROS<T>) [see code], small needle spans are special-cased for T=byte and T=char and sent to specialized helpers in SpanHelpers. But for any other T, these small needle spans are not special-cased, and control flow goes to the generalized SpanHelpers.IndexOfAny<T>(ref T, int, ref T, int) routine, even though n-ary specialized helpers exist.

We should consider special-casing these lengths within the MemoryExtensions.IndexOfAny and LastIndexOfAny methods and funneling to the specialized n-ary helpers. Or, if there's a specific reason this logic was left out for non-byte / non-char types T, we should put a comment in the code to clarify why this logic has been excluded.

[ghost]

Tagging subscribers to this area: @GrabYourPitchforks, @carlossanlop
See info in area-owners.md if you want to be subscribed.

Issue Details

---

Based on a conversation at #53652 (comment).

In the implementation of MemoryExtensions.IndexOfAny<T>(ROS<T>, ROS<T>) [see code], small needle spans are special-cased for T=byte and T=char and sent to specialized helpers in SpanHelpers. But for any other T, these small needle spans are not special-cased, and control flow goes to the generalized SpanHelpers.IndexOfAny<T>(ref T, int, ref T, int) routine, even though n-ary specialized helpers exist.

We should consider special-casing these lengths within the MemoryExtensions.IndexOfAny and LastIndexOfAny methods and funneling to the specialized n-ary helpers. Or, if there's a specific reason this logic was left out for non-byte / non-char types T, we should put a comment in the code to clarify why this logic has been excluded.

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