Improve performance by running initial test and mutating in parallel

[richardwerkman]

Is your feature request related to a problem? Please describe.
Stryker is slow..

Describe the solution you'd like
The current workflow has to wait for the initial testrun to finish to start mutating the syntax trees. This could happen in parallel to improve performance on multi threaded systems.

Additional context
On some large projects mutating can take up seconds, just like the initial testrun. By running this in parallel we can save some time.

[rouke-broersma]

I think this can be safe, but only up to mutant generation. I think we may run into file locking issues if we include compilation.

[richardwerkman]

You're right. I'll see if I can delay writing the mutated assembly to disk. I'm also thinking if this could impact the duration of the initial testrun. Since we will use more resources it could be a bit slower, impacting the timout value we use. But this would be minimal I think..

[dupdob]

Bear in mind that projects mutation is also parallelized, so test and project mutation will compete for ressources, except when the number of projects is low.

[richardwerkman]

@dupdob Yes that is what I meant with use more resources. We could limit the number of threads for the mutating to leave some room for the initial test process. For example we could use the --concurrency setting to limit the threads for mutating. But I'd prefer we just use all the resources we can get to really push performance. And like you said, we use one thread per project. So this would only happen when running on a large solution with many projects.