Forget and Explain: Transparent Verification of GNN Unlearning