During concurrent mark phase, there should not have forwarded objects in roots. Therefore, it should use ShenandoahMarkRefsClosure to mark concurrent roots, instead of ShenandoahMarkResolveRefsClosure.
Note: this is *not* a correctness bug, but performance one, as ShenandoahMarkResolveRefsClosure closure unnecessarily resolves forwarding pointers, where always resolved to themselves.
Note: this is *not* a correctness bug, but performance one, as ShenandoahMarkResolveRefsClosure closure unnecessarily resolves forwarding pointers, where always resolved to themselves.