Woraround for registry resolving when restarting celery prefork workers

[perronld]

Workaround pyramid.threadlocal:get_current_registry() when restarting Celery workers processes (e.g. --max-tasks-per-child)

[github-actions]

Thanks for submitting a PR.
Make sure that you have added tests to validates new features or bugfixes. Also, ensure that existing test suites are still working with the change using the commit test statuses.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (e5f79a5) 85.18% compared to head (d85cf99) 85.19%.

❗ Current head d85cf99 differs from pull request most recent head ab1b834. Consider uploading reports for the commit ab1b834 to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #589      +/-   ##
==========================================
+ Coverage   85.18%   85.19%   +0.01%     
==========================================
  Files          79       79              
  Lines       18023    18026       +3     
  Branches     2728     2729       +1     
==========================================
+ Hits        15353    15358       +5     
+ Misses       1940     1939       -1     
+ Partials      730      729       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[fmigneault]

How does it make a workaround?

The doc seems to indicate that get_current_request is just as bad as get_current_registry.
https://docs.pylonsproject.org/projects/pyramid/en/latest/api/threadlocal.html

Also, what happens if the code is not running with celery (ie: API or CLI), and get_registry() is called. Wouldn't that possibly create an endless recursion?