Asorted tests improvements

[majocha]

Cherrypicked from #17662.

• replace or remove some of the flakiest hardcoded timeouts
• make sure there are no background tasks started with Async.Start etc. and not observed / awaited. They will escape the confinement of currently running test case and if they throw, they bring down the whole test host
• Enabled a few regression tests that were disabled for years because of trivial exit 0 in the code
• named test runs in CI to make test summary easier to use
• added some rudimentary tests for Mailbox.TryScan because we had none outside of the core\controlMailbox script.

[github-actions]

✅ No release notes required

[majocha]

Another one:
https://dev.azure.com/dnceng-public/public/_build/results?buildId=829273&view=logs&j=966a6946-91f0-5397-81f4-d998467880f9&t=344839a9-08c6-5829-65fb-26f32c5a016f&l=5086

 Failed EmittedIL.TryCatch.Stackoverflow reproduction(compilation: StackOverflowRepro.fs) [30 s]
  Error Message:
   System.TimeoutException : Timeout executing command 'dotnet' 'D:\a\_work\1\s\artifacts\bin\FSharp.Compiler.ComponentTests\Release\net9.0\tests\EmittedIL\TryCatch\Stackoverflow_reproduction\StackOverflowRepro.exe'

This comes down to the timeout here being too tight for the CI occasionally:

fsharp/tests/FSharp.Test.Utilities/CompilerAssert.fs

Lines 665 to 666 in 8eec34b

	let timeout = 30000
	let exitCode, output, errors = Commands.executeProcess (Some fileName) arguments (Path.GetDirectoryName(outputFilePath)) timeout

Maybe it would make sense to get rid of such hardcoded timeouts altogether. In the IDE it's obvious when a test hangs and in the CI we can just use --blame-hang-timeout.

[majocha]

I used CancellationTokenSource to cancel MailboxProcessors on test end. Turns out Mailbox uses quite different code paths with and without cancellation, #17849, so this is not acceptable.

[psfinaki]

Thanks for applying proper testing practices, @majocha. This is a really solid effort. I have also learned some multithreading stuff from this :)

[majocha]

Thanks for applying proper testing practices, @majocha. This is a really solid effort. I have also learned some multithreading stuff from this :)

They're still not perfect. Using just sync primitives often obfuscates the code. Now I think some sort of simulated time, Rx style, would make the code more clear. But at the cost of making the tests depend on yet another thing.

[psfinaki]

Well - we're leaving the playground in the better shape we've found it 👀

[Martin521]

Great improvement indeed