Introduce a longer lived 'crew' cache

[niccofyren]

Since most crew based calculations/queries can take quite a while I want to avoid hitting users with this delay as often as possible. We already have a cache in place, but will still hit one user ever 1 hours. This extends this to be only once per week.

Hopefully this shouldn't be a very risky change:

• We already cache these queries, just for a slightly shorter time
• There are a lot of existing cache invalidation code for purging these keys on relevant updates

If we have any instances where purging fails today this will likely make them more visible, but we should probably aim to fix those cases if/when we find any 😉