Adding paraphrasing mining to detect similar tweets

Uses pairwise distance (cosine sim) on the entire dataset embeddings to detect which tweets are similar. Output a list of tweets that are grouped together according to the threshold defined in config.yaml