Exploring Speech Foundation Models for Speaker Diarization in Child-Adult Dyadic Interactions