Update requirements for Python 3.12 support, ignore some warnings

[teutoburg]

To get rid of a DepracationWarning about utcfromtimestamp from a dependency inside matplotlib...

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (7da0dd3) 79.92% compared to head (56c5411) 79.92%.
Report is 3 commits behind head on dev_master.

Additional details and impacted files

@@             Coverage Diff             @@
##           dev_master     #289   +/-   ##
===========================================
  Coverage       79.92%   79.92%           
===========================================
  Files             143      143           
  Lines           14917    14917           
===========================================
  Hits            11922    11922           
  Misses           2995     2995           

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

[teutoburg]

Hmm, it's still there because the warning apparently occurs upon import and not use of dateutils inside matplotlib. This might be hard to get rid of, unless dateutils gets updated or matplotlib decides to ditch it entirely.

Also, the minimum dependency tests can't resolve numpy...

Maybe just ignore that particular warning for now, so things don't fail for no real reason...

[hugobuddel]

Fine with me to explicitly ignore the DeprecationWarning and fix it later. The goal of filterwarnings = "error" is to prevent ignoring deprecation warnings by accident and have them bite us at an inconvenient time. As in, I had to fix ScopeSim because things broke overnight because some functionality in a dependency was removed after years of deprecation.

The idea behind the minimum dependencies test was that it should catch problems with people upgrading ScopeSim, but not e.g. numpy, and then having things break. That logic does not apply here, because Python 3.12 is new. Unfortunately, there doesn't seem to be a version of numpy that both supports Python 3.8 and 3.12, so maybe we should rethink this test.

[hugobuddel]

The minimum dependencies test only runs for Python 3.8, so my comment was invalid. Perhaps we can just upgrade to numpy >= 1.20