This came up while chatting with @ElDeveloper about #237. Currently (unless I'm missing something), empty samples / features are only removed from the table in compress_table(), which is one of the last things done in Empress' python code. It would be preferable to move this work to be earlier in the code so that this is done before tree shearing -- this way, features in the tree and table (but not present in any samples in the table) would be removed from the tree if --p-filter-unobserved-features-from-phylogeny is passed.