Explicit Alignment and Many-to-many Entailment Based Reasoning for Conversational Machine Reading