Split FormatBigInteger into smaller parts.

[kzrnm]

Maybe fix#111708

After investigating the unexpected behavior in #111708, I concluded that JIT optimization is not functioning properly.

By splitting the part of FormatBigInteger that converts bits into base 1,000,000,000 into separate methods, I hope to improve JIT optimization.

This pull request conflicts with my pull request #112178, but I created this as I expect this PR to be backported to .NET 9.

Benchmark

In benchmarks using the released System.Runtime.Numerics.dll, it is about twice as slow compared to \main\corerun.exe.

Code

[MemoryDiagnoser(false)][HideColumns("Job","Error","StdDev","Median","RatioSD")]publicclassToStringTest{[Params([1<<15,1<<20,1<<21,])]publicintN;privatechar[]_dest=newchar[1_000_000];[GlobalSetup]publicvoidSetup(){bigInteger=System.Numerics.BigInteger.One<<N;}BigIntegerbigInteger;[Benchmark(Baseline=true)]publicvoidTryFormat()=>bigInteger.TryFormat(_dest,out_);}

BenchmarkDotNet v0.13.12, Windows 11 (10.0.26100.3037)
13th Gen Intel Core i5-13500, 1 CPU, 20 logical and 14 physical cores
.NET SDK 10.0.100-alpha.1.25077.2
[Host] : .NET 10.0.0 (10.0.25.7313), X64 RyuJIT AVX2
MediumRun : .NET 10.0.0 (42.42.42.42424), X64 RyuJIT AVX2
Job=MediumRun IterationCount=15 LaunchCount=2 WarmupCount=10 

Method	Toolchain	N	Mean	Ratio	Allocated	Alloc Ratio
TryFormat	\main\corerun.exe	32768	985.8 μs	1.00	-	NA
TryFormat	\pr\corerun.exe	32768	984.6 μs	1.00	-	NA
TryFormat	\main\corerun.exe	1048576	1,008,261.6 μs	1.00	-	NA
TryFormat	\pr\corerun.exe	1048576	862,085.5 μs	0.86	-	NA
TryFormat	\main\corerun.exe	2097152	3,957,211.1 μs	1.00	-	NA
TryFormat	\pr\corerun.exe	2097152	3,518,594.2 μs	0.89	-	NA

[kzrnm]

It seems that this might not be the appropriate change.