Use Math.BigMul instead of UInt128.Multiply in BigInteger.Parse

[kzrnm]

UInt128 multiplication includes some unnecessary overhead, so it's better to use Math.BigMul directly.

This PR is a refactored version of #112632.

Benchmark

BenchmarkDotNet v0.16.0-nightly.20260320.467, Windows 11 (10.0.26200.8039/25H2/2025Update/HudsonValley2)
13th Gen Intel Core i5-13500 2.50GHz, 1 CPU, 20 logical and 14 physical cores
Memory: 31.75 GB Total, 11.22 GB Available
.NET SDK 11.0.100-preview.3.26170.106
  [Host]     : .NET 11.0.0 (11.0.0-preview.3.26170.106, 11.0.26.17106), X64 RyuJIT x86-64-v3
  Job-ZICBXN : .NET 11.0.0 (11.0.0-dev, 42.42.42.42424), X64 RyuJIT x86-64-v3
  Job-QENHHP : .NET 11.0.0 (11.0.0-dev, 42.42.42.42424), X64 RyuJIT x86-64-v3

Method	Toolchain	N	Mean	Ratio	Code Size	Gen0	Allocated	Alloc Ratio
PowerOfTenMinusOne	\main\corerun.exe	1000	4.831 μs	1.00	3,285 B	0.0305	440 B	1.00
PowerOfTenMinusOne	\pr\corerun.exe	1000	3.852 μs	0.80	3,284 B	0.0305	440 B	1.00
PowerOfTenMinusOne	\main\corerun.exe	10000	114.386 μs	1.00	3,040 B	0.2441	4184 B	1.00
PowerOfTenMinusOne	\pr\corerun.exe	10000	109.206 μs	0.95	3,040 B	0.2441	4184 B	1.00

<details>

<summary>benchmark code</summary>

```cs
using BenchmarkDotNet.Attributes;
using BenchmarkDotNet.Running;
using BenchmarkDotNet.Configs;
using System.Numerics;
using System.Runtime.InteropServices;

[DisassemblyDiagnoser]
[MemoryDiagnoser(false)]
[HideColumns("Job", "Error", "StdDev", "Median", "RatioSD")]
[GroupBenchmarksBy(BenchmarkLogicalGroupRule.ByMethod)]
public class BigIntParseTest
{
    [Params(1000, 10000)]
    public int N;

    string s1;

    [GlobalSetup]
    public void Setup()
    {
        s1 = new string('9', N);
    }

    [Benchmark] public BigInteger PowerOfTenMinusOne() => BigInteger.Parse(s1);
}

[dotnet-policy-service]

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