Updated numbers

[adamsitnik]

Disclaimer: I don't expect this PR to get merged, I just want to share updated perf numbers to start a discussion.

Empty: an empty console app that prints a boolean and string to the output
EmpyAOT - same app compiled with CoreRT (I've used the 6.0 bits of Microsoft.DotNet.ILCompiler)
Corefxlab - app that uses corefxlab parser (I had to copy-paste the code as the nuget package is not available anymore) to parse a boolean and a string
SystemCommandLineBefore - app that uses System.CommandLine before all recent perf work (PRs)
SystemCommandLineAfter - app that uses System.CommandLine with all recent perf improvements
SystemCommandLineAfterR2R - same as above, but published with PublishReadyToRun=true
SystemCommandLineAfterAOT - same as above, but compiled to native library (possible due to all the @jonsequitur recent work)

Method	Args	Mean	Error	StdDev
Empty	--bool -s test	35.67 ms	1.130 ms	1.301 ms
EmptyAOT	--bool -s test	21.24 ms	0.643 ms	0.714 ms
Corefxlab	--bool -s test	54.72 ms	1.214 ms	1.398 ms
SystemCommandLineBefore	--bool -s test	79.65 ms	1.108 ms	1.232 ms
SystemCommandLineAfter	--bool -s test	60.50 ms	1.123 ms	1.248 ms
SystemCommandLineAfterR2R	--bool -s test	73.21 ms	1.332 ms	1.480 ms
SystemCommandLineAfterAOT	--bool -s test	28.09 ms	0.651 ms	0.697 ms

The initial gap with coreflab parser was 25ms for given scenario. We were able to reduce this gap by 20ms.