Given a feature table and the associated feature sequences, cluster the features against a reference database based on user-specified percent identity threshold of their sequences. Any sequences that don't match are then clustered de novo. This is not a general-purpose clustering method, but rather is intended to be used for clustering the results of quality- filtering/dereplication methods, such as DADA2, or for re-clustering a FeatureTable at a lower percent identity than it was originally clustered at. When a group of features in the input table are clustered into a single feature, the frequency of that single feature in a given sample is the sum of the frequencies of the features that were clustered in that sample. Feature identifiers will be inherited from the centroid feature of each cluster. For features that match a reference sequence, the centroid feature is that reference sequence, so its identifier will become the feature identifier. The clustered_sequences result will contain feature representative sequences that are derived from the sequences input for all features in clustered_table. This will always be the most abundant sequence in the cluster. The new_reference_sequences result will contain the entire reference database, plus feature representative sequences for any de novo features. This is intended to be used as a reference database in subsequent iterations of cluster_features_open_reference, if applicable. See the vsearch documentation for details on how sequence clustering is performed.