Release v0.2.1

• filters split and tee now support ClassificationData as well
• added metadata-from-name filter to extract meta-data from the current input file name
• added inspect filter that allows inspecting data interactively as it passes through the pipeline
• added empty_str_if_none helper method to ldc.text_utils to ensure no None/null values are output with writers
• upgraded seppl to 0.2.2 and switched to using seppl.ClassListerRegistry