[Refactor] Classification 9/n

[SkafteNicki]

What does this PR do?

Continues work on classification refactor #1001
Prior work:

• [Refactor] Classification 8/n #1175
• [Refactor] Classification 7/n #1167
• [Refactor] Classification 6/n #1163
• [Refactor] Classification 5/n #1159
• [Refactor] Classification 4/n #1151
• [Refactor] Classification 3/n #1145
• [Refactor] Classification 2/n #1143
• [Refactor] Classification 1/n #1054

At this point the core implementations are in place and clean up is beginning. This PR takes care of 3 issues:

1. functional auc is deprecated as an metric and moved to utilities as an compute tool instead. Module AUC is completly deprecated and will be removed in future.
2. kl_divergence is deprecated from classification and moved to regression where it better fits (based on the fact that target should be a float tensor for this metric)
3. the default value of average is changed from micro to macro. This is to prevent uses getting the same value from precision, recall and f1 in the multiclass/multilabel cases. Essentially the following holds:

   from torchmetrics.functional import multiclass_precision, multiclass_recall
   import torch
   p, t = torch.randint(3, (20,)), torch.randint(3, (20,))
   multiclass_precision(p, t, num_classes=3, average="micro") == multiclass_recall(p, t, num_classes=3, average="micro")
   # tensor(True)
   multiclass_precision(p, t, num_classes=3, average="macro") == multiclass_recall(p, t, num_classes=3, average="macro")
   # tensor(False)

   it is just not wrong to use micro in these cases, the underlying statistics just collapse into the same numbers in these cases. This change should hopefully prevent such confusion for users.

Before submitting

• Was this discussed/approved via a Github issue? (no need for typos and docs improvements)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure to update the docs?
• Did you write any new necessary tests?

PR review

Anyone in the community is free to review the PR once the tests have passed.
If we didn't discuss your PR in Github issues there's a high chance it will not be merged.

Did you have fun?

Make sure you had fun coding 🙃

[codecov]

Codecov Report

Merging #1189 (6875845) into devel/classification (031ffdd) will decrease coverage by 0%.
The diff coverage is 79%.

@@                 Coverage Diff                  @@
##           devel/classification   #1189   +/-   ##
====================================================
- Coverage                    89%     89%   -0%     
====================================================
  Files                       189     191    +2     
  Lines                     10767   10786   +19     
====================================================
- Hits                       9587    9577   -10     
- Misses                     1180    1209   +29     

[justusschock]

We probably don't strictly need to follow update-compute in our utils, especially if there is not modular interface so I propose a slightly more descriptive name :)

Other than that: LGTM as always :)