[tests] Enable sys cppia tests

[tobil4sk]

There were failures due to: HaxeFoundation/hxcpp#901. I've added workarounds for now to avoid the errors.

[tobil4sk]

This threads test currently works incorrectly in jit mode:

haxe/tests/threads/src/cases/TestEvents.hx

Lines 35 to 38 in ffd3b45

	raises(
	() -> thread.events.run(() -> {}),
	sys.thread.NoEventLoopException
	);

This is because local functions break exceptions in jit mode, see: HaxeFoundation/hxcpp#876 (comment)

[Simn]

The hxcpp PR has been merged, please update the branch.

[tobil4sk]

There are still cppia jit failures. I think it is due to: HaxeFoundation/hxcpp#1212, but it's strange that it isn't affecting Linux because the minimal sample does.

[tobil4sk]

it's strange that it isn't affecting Linux because the minimal sample does.

I've found that on linux there is an element of randomness, the same code can cause the issue or not depending on where it is found in the program and whether the client script was compiled with export_classes.info (which changes the classes included/excluded from the cppia script).

I think the problem is that there is some out of bounds memory access occurring due to the attempt to run a method on a null object, and in some cases this out of bounds access is not detected on Linux, but is always detected on Windows and Mac.

Either way, the hxcpp patch includes some additional tests and fixes these errors.

[tobil4sk]

@Simn The cppia sys tests are working now. We just need to fix the cppia host path for arm64: HaxeFoundation/hxcpp#1323. It would also be useful to clean up some linux arm64 warnings in the cppia host build: HaxeFoundation/hxcpp#1282

I have reverted the thread/eventloop change as those tests are currently broken due to a regression: HaxeFoundation/hxcpp#1321. I will open a separate PR for them once this is merged.