Watch and evaluate performance impact of changes

[JanKrivanek]

Context

When analyzers are disabled ('/analyze' not passed on command line) there should be no to minimal impact on the build perf. We should very closely watch and resolve any dfeviations from that. By the current PerfStar runs it seems there are currently gaps in this way already.

Once we resolve issues with the 'no analysis perf baseline', we should setup the 'analysis opted in perf baseline' - have a perfstar runs with the '/analyze' opt-in. We should evaluate reasons for possible deviations in this mode from the 'no analysis perf baseline' and as well track the changes between the main versions.

Goals

• [PerfStar] Add tests for BuildCheck #10828
• Check performance with/without analyzers. #9988
• Investigate the source of a big perf overhead when using analyzers on big solutions. #10068
• https://github.com/orgs/dotnet/projects/117/views/22?pane=issue&itemId=66221323
• Turning on PerfStar https://dev.azure.com/devdiv/DevDiv/_workitems/edit/2057021