Automatic Data Augmentation Selection and Parametrization in Contrastive Self-Supervised Speech Representation Learning