Re-contextualizing Fairness in NLP: The Case of India