Deal nicely with swh-storage's backend down time in clients
At the moment, 500 errors pops up whenever the db or the swh-storage api is down.
Deal more nicely with this, typically by returning a 'standard' (as per now it's like that): {'error': 'blah blah something went wrong blah blah'}
with the nice 503 errors
Migrated from T269 (view on Phabricator)