Drive feature matrix from a file instead of database queries

Storign the data in the file both puts it under git control, and it
allows new entries to be staged when a new release is made, to be
automatically pushed alongside the other release entries.

As a bonus, versions now only need to be specified when the state
changes, not with a massive set of checkboxes, one for each version
since forever.

Imports all the existing content from the current feature matrix for
today. Leaves the database tables in place for now in case we need to
back out, but they should be removed soon. For now, just remove them
from /admin/ so nobody mistakenly tries to update them.

Finally, changes the URLs for feature details to be slug-based instead
of id-based, and keeps the old id mapping around for redirects.

Author: mhagander