More aggressive ahead of time optimization for string-key type formatter.

[pCYSl5EDgo]

Summary

Current UniversalCodeGenerator generated formatter code unnecessarily allocates byte[] array and others. This pull request reduces memory allocation, improves runtime initialization performance and embeds specialized automaton code for deserialization.

Benchmark

BenchmarkDotNet=v0.12.0, OS=Windows 10.0.18363
Intel Core i7-8750H CPU 2.20GHz (Coffee Lake), 1 CPU, 12 logical and 6 physical cores
.NET Core SDK=3.1.200
  [Host]     : .NET Core 3.1.2 (CoreCLR 4.700.20.6602, CoreFX 4.700.20.6702), X64 RyuJIT
  Job-RCOQVZ : .NET Core 3.1.2 (CoreCLR 4.700.20.6602, CoreFX 4.700.20.6702), X64 RyuJIT

Runtime=.NET Core 3.1  

Method	ConstantString	Mean	Error	StdDev	Median
DeserializeTest_NewEmbeddedFormatter		669.1 ns	13.12 ns	13.48 ns	667.4 ns
DeserializeTest2_OriginalGeneratedFormatter		924.4 ns	18.56 ns	42.64 ns	914.6 ns
DeserializeTest3_DynamicILGenerator_Formatter		675.5 ns	13.49 ns	24.66 ns	672.1 ns
DeserializeTest_NewEmbeddedFormatter	a	751.3 ns	21.10 ns	60.21 ns	731.5 ns
DeserializeTest2_OriginalGeneratedFormatter	a	930.9 ns	15.06 ns	14.09 ns	929.3 ns
DeserializeTest3_DynamicILGenerator_Formatter	a	672.3 ns	13.36 ns	13.72 ns	666.1 ns
DeserializeTest_NewEmbeddedFormatter	shorter	736.6 ns	16.93 ns	15.83 ns	731.6 ns
DeserializeTest2_OriginalGeneratedFormatter	shorter	1,005.6 ns	19.96 ns	48.20 ns	995.6 ns
DeserializeTest3_DynamicILGenerator_Formatter	shorter	694.8 ns	12.88 ns	12.65 ns	692.0 ns
DeserializeTest_NewEmbeddedFormatter	このベンチ(...) [62]	1,208.8 ns	22.94 ns	24.54 ns	1,206.4 ns
DeserializeTest2_OriginalGeneratedFormatter	このベンチ(...) [62]	1,550.3 ns	45.38 ns	128.74 ns	1,510.0 ns
DeserializeTest3_DynamicILGenerator_Formatter	このベンチ(...) [62]	1,169.7 ns	16.99 ns	14.19 ns	1,165.6 ns

Benchmark Project Gist.