🐛 Fix MNIST test dataloader for shifted data

[tonyzamyatin]

Testing on the MNISTDataModule inside a ClassifcationRoutine with eval_shifted=True set would raise a ValueError("No samples to concatenate") when self.test_shift_metrics.compute() was called inside the ClassificationRoutine, because test_dataloader() did not actually return the test dataloader with the shifter data.

Also, MNISTC was missing the property shift_severity which is referenced inside ClassificationRoutine.

[o-laurent]

Thanks for your PR! I'll make the changes highlighted in the comments.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.16%. Comparing base (7d0d345) to head (d4d09f5).
Report is 10 commits behind head on dev.

Additional details and impacted files

@@           Coverage Diff           @@
##              dev     #135   +/-   ##
=======================================
  Coverage   99.15%   99.16%           
=======================================
  Files         143      143           
  Lines        7136     7145    +9     
  Branches      910      912    +2     
=======================================
+ Hits         7076     7085    +9     
  Misses         27       27           
  Partials       33       33           

Flag	Coverage Δ
cpu	99.16% <100.00%> (+<0.01%)	⬆️
pytest	99.16% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

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