Improve ProbabilisticMap performance for small value sets

[MihaZupan]

Since IndexOfAnyExcept does a value.Contains(c) on every input character, it's faster if the value is at least 8 characters long.

We could go further with e.g. working with a Vector128<ushort> directly instead of calling into Contains, but I don't know how common short inputs will actually be in practice.

Method	Toolchain	Length	Mean	Error	Ratio
IndexOfAnyExcept6	main	10000	82.06 us	0.288 us	1.00
IndexOfAnyExcept6	pr	10000	31.00 us	0.491 us	0.38

[ghost]

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

Issue Details

---

Since IndexOfAnyExcept does a value.Contains(c) on every input character, it's faster if the value is at least 8 characters long.

We could go further with e.g. working with a Vector128<ushort> directly instead of calling into Contains, but I don't know how common short inputs will actually be in practice.

Method	Toolchain	Length	Mean	Error	Ratio
IndexOfAnyExcept6	main	10000	82.06 us	0.288 us	1.00
IndexOfAnyExcept6	pr	10000	31.00 us	0.491 us	0.38

Author:	MihaZupan
Assignees:	MihaZupan
Labels:	area-System.Buffers
Milestone:	8.0.0

[stephentoub]

Clever.

[MihaZupan]

Build failure is known according to Build Analysis