Small GenerateOperationId optimization

[Inok]

Hi!

I noticed that GenerateOperationId is called in almost every operation on the cache. I optimized the implementation a bit:

• Eliminated boundary checks on the char[] dictionary (replacing to const string helps the compiler to figure out that boundary checks are not needed).
• Switched from ThreadLocal buffer to stackalloc char[13] on netstandard2.0 (slightly cheaper and has no boundary checks as well).
• Switched to modern string.Create() on net8.0+ to build the string in place rather than build a buffer and then copy it to the string.

Results:
netstandard2.0 version became slightly faster, but no big difference.
net8.0 version is ~30% faster.

Method	Mean	Error	StdDev	Gen0	Gen1	Allocated
GenerateOperationId_Original	15.69 ns	0.221 ns	0.207 ns	0.0038	0.0000	48 B
GenerateOperationId_NetStandard	14.61 ns	0.168 ns	0.158 ns	0.0038	0.0000	48 B
GenerateOperationId_Net8Plus	10.35 ns	0.198 ns	0.186 ns	0.0038	0.0000	48 B

@jodydonetti what do you think?

[Inok]

And I also noticed that AOTTester targets net10, but global.json and CI pipeline use .NET 9 SDK, which breaks the build. I bumped SDK to .NET 10 for now, if you don't mind 🙂

[jodydonetti]

Hi @Inok and thanks for the PR!

Will look at it in the next few days, but it already looks good.