After a bit of pondering - this will be very hard to analyse. Comparing direct numbers is not really possible, as the download numbers do not reflect the ratio of users that have extras-devel enabled on a regular basis (e.g. if you install a version that is present in Extras, that will count as extras even if you have extras-devel enabled, but error is possible the othey way, too - just because someone turned extras-devel on for a single package doesn't mean he keeps it turned on constantly).