Use BitVector128 in CharacterMap<T>

[KrisVandermotten]

CharacterMap<T> uses a bool[] isOpeningCharacter for quick lookup.

This PR replaces that by a custom BitVector128 that stores 128 bits is two ulong fields using a fixed size buffer.

This results in a measurable speedup:

The benchmark:

public class Program
{
    private string text = File.ReadAllText("spec.md");

    [Benchmark]
    public void TestMarkdig()
    {
        Markdown.Parse(text);
    }

    private static void Main()
    {
        BenchmarkRunner.Run<Program>();
    }
}

The results on .NET Framework:

BenchmarkDotNet=v0.10.6, OS=Windows 10.0.18363
Processor=Intel Core i7-7700 CPU 3.60GHz (Kaby Lake), ProcessorCount=8
Frequency=10000000 Hz, Resolution=100.0000 ns, Timer=UNKNOWN
  [Host]     : Clr 4.0.30319.42000, 64bit RyuJIT-v4.8.4075.0
  DefaultJob : Clr 4.0.30319.42000, 64bit RyuJIT-v4.8.4075.0

Before:

Method	Mean	Error	StdDev
TestMarkdig	3.363 ms	0.0209 ms	0.0196 ms

After:

Method	Mean	Error	StdDev
TestMarkdig	3.275 ms	0.0220 ms	0.0206 ms

The results on .NET Core:

BenchmarkDotNet=v0.12.0, OS=Windows 10.0.18363
Intel Core i7-7700 CPU 3.60GHz (Kaby Lake), 1 CPU, 8 logical and 4 physical cores
.NET Core SDK=3.1.101
  [Host]     : .NET Core 3.1.1 (CoreCLR 4.700.19.60701, CoreFX 4.700.19.60801), X64 RyuJIT
  DefaultJob : .NET Core 3.1.1 (CoreCLR 4.700.19.60701, CoreFX 4.700.19.60801), X64 RyuJIT

Before:

Method	Mean	Error	StdDev
TestMarkdig	3.128 ms	0.0207 ms	0.0193 ms

After:

Method	Mean	Error	StdDev
TestMarkdig	3.034 ms	0.0173 ms	0.0145 ms

Conclusion: parsing time is reduced by about 3 %.

[KrisVandermotten]

I changed the implementation to use four uint values instead of two ulong values.

The performance is identical on X64, but it can only be better on X86 or other 32 bit platforms.

[xoofx]

Nice trick, thanks!