DELULU: Discriminative Embedding Learning Using Latent Units for Speaker-Aware Self-Supervised Speech Foundational Model