test: re-enable or replace test_update when possible

[terriko]

In #5217 I've disabled most of the parts of test update that actually, well, update, because they're causing out of memory problems.

This is one of few tests we have that actually runs a full update and gives us a warning if one of our data sources is causing problems, although we also have daily cache update jobs so it's not like we wouldn't notice. Unfortunately, any test relying on network stuff is a problem, so we could probably stand to re-think how this test is done and see if we can replace or supplement it with something that works consistently but will trigger the appropriate code paths.

[ffontaine]

Perhaps this test_update could be re-enabled after merging #4956