Fail Faster on Errors

We've had a couple of problems where something wasn't quite working as expected, but there was an error or warning in the logs that could have tipped us off sooner:

https://github.com/dotnet/performance/issues/2186
https://github.com/dotnet/BenchmarkDotNet/issues/1887

This is an issue to propose better mechanisms of failling fast in these cases, so problems are detected sooner.

I can think of a few ways to do this:

1. Changes to BDN (could work in some cases)
2. Parsing logs
3. `Benchmarks` that are designed to fail if run under the wrong runtime parameters