The patch looks technically good, though I’m not a big fan of the double underscores, like element_get__text(). Maybe element_text_getter() instead, because it has to match the “getter” signature?
Personally, I would say keep the doc strings basic. If you want to add more depth about “tail”, I suggest to keep it to the main RST documentation (which was updated fairly recently). But it would be good to quickly distinguish text and tail other than by name. Maybe:
{"text", ..., "A string of text directly after the start tag, or None"},
{"tail", ..., "A string of text directly after the end tag, or None"},