This breaks translation processing; multi-domain support needs to be added to
Flask-Babel. The changes made here put the "important" strings that are used
most often into the priority domain, and split the content from different
sub-urls into separate domains. Multi-domain support will be implemented such
that translations for a specific request are pulled both from the domain which
corresponds to the request url and the priority domain.
This does mean that there may be duplication of translation strings between
domains, but that is unavoidable (and the translators prefer working with
smaller resource files).
The jinja2 extracter needs to know about our custom HighlightExtension (used on
pages such as get-involved/develop/applications), which is not in the env's
python path. Copying pybabel into the base dir gives it access to our modules.
This is only required for string extraction; other scripts use pybabel from the
env directly.
If the env's pybabel version is updated, this copy must also be updated.
