I think wcwidth should drop all but the latest unicode tables, the unicode_version argument is marked deprecated and we should just ignore it, and only generate and use the latest versions of all multi-version tables (zero and wide).

all downstream dependencies have no interest in supporting many versions, the only ones are my own, like in ucs-detect so I can manage to resolve that dependency. I'm about half-way through total analysis of medium-large dependencies so far, I don't expect to find any use or need.

We should still {"version key": (..)} them, and keep the generated code format of the tables, but they will be one item long, with their version number for compatibility and list_versions() returning only the one version and so on.