add RecordParser

[leandromoh]

Hi! I saw your post and I would like to add my parser on benchmark, when this PR is finished.

[joelverhagen]

Hey @leandromoh, thanks for taking the time to open a PR. It looks like the CI is complaining about some missing types. Perhaps we're missing the package reference in the .csproj to https://www.nuget.org/packages/recordparser?

[leandromoh]

Perhaps we're missing the package reference in the .csproj

yes, I was working with a local reference, whilist the PR was in draft, to do some experiments..

It was a bit funny that your benchmark test the exactly opposite cenario that my lib was thought to resolve: when you want to get all column values as string before to parse to specific type (ex. int, DateTime, etc). I designed RecordParser exactly to avoid intermediary unnecessary string allocations, if a field is some value type, like int for example, I parse it without intermediary string, in a straight manner from ReadOnlySpan<byte> (from file) to ReadOnlySpan<char> and finally int. This way avoids unnecessary pressure for GC / memory allocations.

Fortunately the lib had a good result even in this unpredicted cenario, at least on my pc.

BenchmarkDotNet=v0.13.0, OS=Windows 10.0.18363.1440 (1909/November2019Update/19H2)
Intel Core i7-8650U CPU 1.90GHz (Kaby Lake R), 1 CPU, 8 logical and 4 physical cores
.NET SDK=5.0.104
  [Host]     : .NET 5.0.4 (5.0.421.11614), X64 RyuJIT
  DefaultJob : .NET 5.0.4 (5.0.421.11614), X64 RyuJIT

Method	LineCount	Mean	Error	StdDev	Median	Gen 0	Gen 1	Gen 2	Allocated
RecordParser	1000000	2.978 s	0.0591 s	0.1450 s	2.936 s	58000.0000	21000.0000	3000.0000	345 MB
CsvHelper	1000000	3.546 s	0.4463 s	1.2293 s	2.960 s	44000.0000	17000.0000	3000.0000	261 MB
Cursively	1000000	2.214 s	0.0406 s	0.0733 s	2.191 s	58000.0000	21000.0000	3000.0000	345 MB
Sylvan_Data_Csv	1000000	2.911 s	0.4912 s	1.4252 s	2.088 s	44000.0000	17000.0000	3000.0000	261 MB

[leandromoh]

@joelverhagen PR is ready for review