Exploration of Language Dependency for Japanese Self-Supervised Speech Representation Models