dotnet/runtime#1506 optimize huffman decoding

[rokonec]

Optimized to use 8 bits lookup tables tree.

dotnet/runtime#1506

Shared with runtime. Already reviewed by @Tratcher in PR dotnet/runtime#43603

Fixes #18943

[BrennanConroy]

Bot opened the same PR #28242

[Tratcher]

I thought you were going to send this PR and run the benchmarks before merging the change in runtime? Do we still have benchmarks we need to run here?

I'll close #28242 since it's redundant.

[rokonec]

@Tratcher benchmarks has been created and run in https://github.com/dotnet/runtime/blob/master/src/libraries/System.Net.Http/tests/PerformanceTests/HPackHuffmanBenchmark/Program.cs

Latest version of comparison benchmarked had following results:

Method	Mean	Error	StdDev	Op/s	Ratio
Old	2,146.14 ns	40.559 ns	39.835 ns	465,953.3	1.00
New	537.22 ns	2.607 ns	2.439 ns	1,861,447.5	0.25

Unfortunately

aspnetcore/src/Servers/Kestrel/perf/Kestrel.Performance/HPackDecoderBenchmark.cs

Line 47 in 47138f9

static HPackDecoderBenchmark()

do not contain benchmarks for huffman encoded headers.

Shall I add it there?

[Tratcher]

Ok. I don't know if we need duplication of those benchmarks. @halter73 do you care?

[halter73]

I'm fine with not duplicating. Maybe we should remove the HPACK microbenchmarks from Kestrel if they're covered by the runtime so we don't waste time on the Kestrel versions.