test: make StdlibUnittest.CrashingTests pass on Windows

[compnerd]

The output can be interleaved differently. Accept either ordering.

Replace this paragraph with a description of your changes and rationale. Provide links to external references/discussions if appropriate.

Resolves SR-NNNN.

[compnerd]

@swift-ci please test

[compnerd]

CC: @kubamracek

[swift-ci]

Build failed
Swift Test OS X Platform
Git Sha - 4c91a3e

[compnerd]

@swift-ci please test macOS platform

[swift-ci]

Build failed
Swift Test OS X Platform
Git Sha - 4c91a3e

[compnerd]

@swift-ci please test macOS platform

[swift-ci]

Build failed
Swift Test OS X Platform
Git Sha - 4c91a3e

[compnerd]

sigh lldb again

[compnerd]

@swift-ci please smoke test macOS platform

[compnerd]

@swift-ci please test macOS platform

[swift-ci]

Build failed
Swift Test OS X Platform
Git Sha - 4c91a3e

[compnerd]

@swift-ci please test macOS platform

[swift-ci]

Build failed
Swift Test OS X Platform
Git Sha - 4c91a3e

[drodriguez]

So I think I have some idea why this one was needed and why is failing again in Azure in another of the test. For all the platforms except Windows, the stdlib unit test framework uses select to read from the pipes connected to the stdout and stderr of the test processes. select should maintain the relative order of the messages generated by the test process. However, in Windows, we spin two threads that read from the handles "at the same time", which will not maintain the relative order.

I tried to play a little and enable using select also in Windows, but I got into a lot of dead ends. I wonder if there's a better alternative to the two threads in Windows. Something that can maintain the relative order of the events without leaving it to the scheduler.

Adding DAG to more cases is not only a nuisance, but also wrong, because output happens before error, but the test is saying that it doesn't matter.