Benchmarking Children's ASR with Supervised and Self-supervised Speech Foundation Models