Introducing proper deallocation ways for CPPInstance

[keremsahn]

Built upon #37, just last commit is related to this PR
I added three flags to the CPPInstance to distinguish different allocation methods, and in HandleReturn ,where ownership is given, corresponding flags are set after analyzer/attribute check.

Using these new flags CPPInstance::op_dealloc_nofree introduces different kind of deallocations.
Also tests changed a lot after this feature

And I need to explain why I added kIsArrayAlloc instead of using kIsArray, users can set kIsArray flag by reshape feature:

Klass* alloc(size_t count){  return ::operator new(sizeof(Klass)*count);}

and if user reshapes the returned object, that will lead UB because object is deallocated by operator delete[]

@Vipul-Cariappa @aaronj0 @vgvassilev

[keremsahn]

I do not have access to a mac and I am trying to figure out why macos tests fail, that is why I am pushing constantly

[Vipul-Cariappa]

Suggested change

	alloc()
	assert cppjit.gbl.memory.memOwn.dtorCount == 0
	alloc()
	gc.collect()
	assert cppjit.gbl.memory.memOwn.dtorCount == 0

[aaronj0]

Heads-up: main was force-pushed before the first release as a one-time fix for authorship on two commits ported from the cppyy forks, which credited claude instead of the original patch authors. Since this branch was based on the old history, I rebased it onto the current main and force-pushed, your commits are unchanged. If you have local work on this branch, fetch and put it on top with git rebase <your-remote>/<branch> (or git reset --hard <your-remote>/<branch> if your local copy has nothing extra).