Use .NET 5's HEX function

[kiminuo]

This PR aligns us with .NET and uses Convert.FromHexString which is vectorized in .NET, so we should observe significant speedup (see MetacoSA/NBitcoin#1149 for some numbers).

Current master implementation

|  Method |      Mean |    Error |   StdDev |
|-------- |----------:|---------:|---------:|
|   ToHex |  25.19 ns | 0.501 ns | 1.355 ns |
| FromHex | 563.06 ns | 4.742 ns | 3.702 ns |

This PR: Only first commit:

|  Method |     Mean |    Error |   StdDev |
|-------- |---------:|---------:|---------:|
|   ToHex | 35.58 ns | 0.171 ns | 0.152 ns |
| FromHex | 71.38 ns | 0.650 ns | 0.576 ns |

This PR: All commits

|  Method |     Mean |    Error |   StdDev |
|-------- |---------:|---------:|---------:|
|   ToHex | 25.80 ns | 0.474 ns | 1.180 ns |
| FromHex | 75.86 ns | 1.552 ns | 1.963 ns |

Benchmark is here, use in WalletWasabi.Tests folder:

dotnet run -c Release --framework net6.0 -- --runtimes net6.0

With this PR, we can replace https://github.com/zkSNACKs/WalletWasabi/blob/master/WalletWasabi/Backend/Models/FilterModel.cs#L47 to use our new implementation and it can shave a bit of time from startup. Fundamentally, we should tackle it differently, but given current consensus, this is what we can do.

[turbolay]

ByteHelpers.FromHex(): 0.01162376 ms / read a compact filter
Encoders.Hex.DecodeData(): 0.01115517 ms / read a compact filter

-> No difference for the main operation that is using this because the NBitcoin implementation is already fast.
Yet, it's more compact and faster that the code we currently have.