Use the standard thread library for the lua interpreter

[TheGondos]

This PR replaces Windows specific thread handling with standard C++ constructs in the lua interpreter/console/MFD.
I've tested it on the DG atmospheric autopilot and solar system scenarios but I'd prefer to have some more feedback.
Is there some kind of stress test for the lua interpreter?

[TheGondos]

Working on the failed test case, I think there is a bug in the core that is worked around with a TerminateThread,..

[dimitry-ishenko]

@TheGondos did you see my comments in your _itoa PR?

[TheGondos]

@TheGondos did you see my comments in your _itoa PR?

Yes, I can work around the 32bit warning but I think the root cause is using an INT_PTR to represent either an ExternMFD pointer or an MFD Id. Maybe it would be better to use a std::variant instead (I replaced it with a struct containing a union on my linux branch)

[dimitry-ishenko]

Sorry, since that PR was accepted I wasn't sure if you saw my comments. Let's continue in there.

[TheGondos]

Since we are now using C++20, I took the liberty to switch to semaphores since they are more appropriate.
There was a deadlock when calling the Interpreter destructor in case its thread was already finished when trying to step it.
There was also a crash when calling exit from the intepreter thread so it now forwards its exit code to the main thread and exit is called from there.
CTests are now OK, I tried the DG atmo autopilot scenario several times in a raw and got no crash/freeze when exiting the scenario.
I'd still prefer to have some feedback before marking this ready to merge though.

[dimitry-ishenko]

All those late nights I've spent updating the code for c++20 are starting to pay off already 😅

[TheGondos]

BTW, I tend to rebase/squash before merging because I find it easier to revert/cherry-pick if needed later. Tell me if it's not something you want.

[Xyon]

BTW, I tend to rebase/squash before merging because I find it easier to revert/cherry-pick if needed later. Tell me if it's not something you want.

Nah it's fine, I generally prefer to do that anyway unless the history on the branch is significant somehow

[TheGondos]

Added a minor prototype change.
BTW, is there a way to run CTests in a loop? It's not really enough to declare some multithreading stuff OK over just a couple of tests. If I could launch it a few thousand times, it would be better...

[DarkWanderer]

Added a minor prototype change. BTW, is there a way to run CTests in a loop? It's not really enough to declare some multithreading stuff OK over just a couple of tests. If I could launch it a few thousand times, it would be better...

https://cmake.org/cmake/help/latest/manual/ctest.1.html#cmdoption-ctest-repeat

ctest --repeat-until-fail 1000 <other parameters>