Mitigating Intra-Speaker Variability in Diarization with Style-Controllable Speech Augmentation