add ignore_index option to exact_match metric

[ItamarChinn]

Fixes 1508

This fix sets preds to equal the ignore_index in the same indices that target == ignore_index such that if all other elements in preds except the ones at indices where target == ignore_index are identical this constitutes an exact match.

Before submitting

• Was this discussed/agreed 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? -> unnecessary, the issue is that the docs claim a functionality that doesn't exist
• Did you write any new necessary tests? -> added minimal test in the comments

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.

The following test checks that it functions correctly:

import torch
import torchmetrics

preds = torch.tensor([[0,0,0,1,1,1,1], [1,1,1,1,1,1,1]])
target = torch.tensor([[0,0,0,1,-100,-100,-100], [0,0,1,1,-100,-100,-100]])
exact_match = torchmetrics.classification.MulticlassExactMatch(task='multiclass', num_classes=2, ignore_index=-100)
exact_match.update(preds, target)
assert exact_match.compute() == torch.tensor(0.5), "MulticlassExactMatch behaving unexpectedly" 

[codecov]

Codecov Report

Merging #1540 (1b06756) into master (b3915a3) will decrease coverage by 36%.
The diff coverage is 100%.

Additional details and impacted files

@@           Coverage Diff            @@
##           master   #1540     +/-   ##
========================================
- Coverage      87%     51%    -36%     
========================================
  Files         216     216             
  Lines       11331   11332      +1     
========================================
- Hits         9842    5752   -4090     
- Misses       1489    5580   +4091