ConcurrentChain: Optimize Load method

[Carti-it]

This PR optimizes ConcurrentChain(byte[] bytes) method by removing LINQ use and using a simple while loop. On my machine, it translates to ~14% improvement when using the constructor.

Benchmark

I started Wasabi (WalletWasabi/WalletWasabi@b28faaf) and it generates WalletWasabi/Client/BitcoinP2pNetwork/BlockHeadersMain.dat (~104 MB) and used that file for benchmarking purposes (benchmark was added in 920bb36 and removed in 7116978):

using BenchmarkDotNet.Attributes;
using System;
using System.IO;

namespace NBitcoin.Bench;

public class ConcurrentChainBench
{
	byte[] Bytes; 

	[GlobalSetup]
	public void Setup()
	{
		var appDataPath = Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.ApplicationData), "WalletWasabi", "Client", "BitcoinP2pNetwork", "BlockHeadersMain.dat");
		Bytes = File.ReadAllBytes(appDataPath);
	}

	[Benchmark]
	public void WriteToString()
	{
		var chain = new ConcurrentChain(Bytes, Network.Main.Consensus.ConsensusFactory);
	}
}

cd NBitcoin.Bench
dotnet run -c Release -- --runtimes net8.0 --filter *ConcurrentChainBench*

master:

Method	Mean	Error	StdDev
WriteToString	1.507 s	0.0298 s	0.0306 s

PR:

Method	Mean	Error	StdDev
WriteToString	1.299 s	0.0214 s	0.0189 s

So it's about 200 ms faster, or 14% faster.

In a way, I believe that Wasabi should optimize the header chain loading differently. However, given I spent some time with this, I figured it may be worth a PR.

[NicolasDorier]

Checking... but I think that people should avoid having to get the ConcurrentChain, because it is taking way too much memory in general.