Text this: Contrastive Speaker Representation Learning with Hard Negative Sampling for Speaker Recognition