Avoid extra allocations in YamlScalarNode GetHashCode

[lahma]

YamlScalarNode calls CombineHashCodes(Tag, Value) which is the only call in library that chooses the params array method. No need to allocate array for this so changed method signature to take the second object as parameter.

EDIT

Then found out that the first parameter is actuall struct (TagName Tag) so causes boxing, is wiser to call hash code directly and now remove the CombineHashCodes overload that is no longer needed at all.

Also marked TagName as read-only struct. Should also use ThrowHelper here, but didn't duplicate introduction work from other PR.

Benchmark Code

Uses the large YAML from issue #519 . I noticed you don't have a benchmark project currently so I'm running my local one.

[MemoryDiagnoser]
public class YamlStreamBenchmark
{
    private string yamlString = "";

    [GlobalSetup]
    public void Setup()
    {
        yamlString = File.ReadAllText(@"c:\work\saltern.yml");
    }

    [Benchmark]
    public void Load()
    {
        var yamlStream = new YamlStream();
        yamlStream.Load(new StringReader(yamlString));
    }
}

Results

BenchmarkDotNet=v0.13.1, OS=Windows 10.0.22000
AMD Ryzen 9 5950X, 1 CPU, 32 logical and 16 physical cores
.NET SDK=6.0.100
  [Host]     : .NET 6.0.0 (6.0.21.52210), X64 RyuJIT
  DefaultJob : .NET 6.0.0 (6.0.21.52210), X64 RyuJIT

YamlStreamBenchmark

Diff	Method	Mean	Error	Gen 0	Gen 1	Gen 2	Allocated
Old	Load	96.60 ms	1.183 ms	5833.3333	2833.3333	1000.0000	81 MB
New		95.72 ms (-1%)	1.659 ms	5666.6667 (-3%)	2500.0000 (-12%)	1000.0000 (0%)	78 MB (-4%)

[aaubry]

This feature has been released in version 12.0.0.