Recommendation or Discrimination?: Quantifying Distribution Parity in Information Retrieval Systems