Fix N+1 query in event calendar found by sentry

[raphaelm]

This fixes a performance issue introduced by the new settings lookup in #4036

[codecov]

Codecov Report

Attention: Patch coverage is 50.00000% with 4 lines in your changes are missing coverage. Please review.

Project coverage is 78.08%. Comparing base (fb403da) to head (4c72b40).
Report is 32 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #4104   +/-   ##
=======================================
  Coverage   78.07%   78.08%           
=======================================
  Files         429      429           
  Lines       60227    60272   +45     
=======================================
+ Hits        47023    47061   +38     
- Misses      13204    13211    +7     

Files	Coverage Δ
src/pretix/base/services/quotas.py	96.61% <100.00%> (+0.01%)	⬆️
src/pretix/presale/views/organizer.py	58.36% <42.85%> (-0.19%)	⬇️

... and 19 files with indirect coverage changes

[luelista]

why is this necessary? shouldn't this already be filled in by the .select_related('event', 'subevent') in EventMixin.annotated (i see that it isn't, but i don't understand why)?

[raphaelm]

Even if it would, select_related would give it a new instance of the Event() class. We want, if possible, to use one instance of the Event() class so we only need to fetch the hierarkey data from cache once :)

[raphaelm]

(That's one reason why we sometimes use prefetch_related where select_related would be sufficient, but also why we do some hacks like this here…)