Attempting to add debug logs for ENQUEUING an invalid object

[NHDaly]

We were seeing GC Corruption when popping an invalid object, but we don't know how we got this invalid pointer. (And we can't repro on linux so can't use RR.)

This is an attempt to check for the object's validity before enqueuing so that we can hopefully give a more useful error message (which object's pointer was corrupted).

But for some reason it's currently complaining about the very first push....

[d-netto]

Fixed some compiler warnings and inserted the gc_assert_parent_validity in the places where the parent validity should be checked when enqueueing an object.

On the example from #21590 (with GC_ASSERT_PARENT_VALIDITY defined):

a = Ref(1)
ptr = pointer_from_objref(a) - 8
unsafe_store!(Ptr{Int}(ptr), ptr)
GC.gc()

GC error (probable corruption)
Allocations: 15177 (Pool: 15157; Big: 20); GC: 0
Parent 0x11809dba0
of type:
Core.Binding
While marking child at 0x1180f4350
of type:
Core.ReturnNode(val=SSAValue(1))

The performance of keeping GC_ASSERT_PARENT_VALIDITY enabled needs to be further assessed in order to justify the decision of keeping under a debug flag or enabling it by default.

[NHDaly]

Awesome. Thanks again for picking this up and carrying it forward, @d-netto!! :)

The hardest part to me will be ensuring that we don't ever miss a spot and push an object without including this assertion there. (Whereas before, when it was on the pop side, it was easier to ensure it always happens.) But I think it should be pretty doable.

Thanks!