Memory optimization of perplexity metric

[nsmlzl]

What does this PR do?

Fixes #2337 ; reduces the memory footprint when computing the perplexity by not creating a huge temporary tensor with n^2 elements (n: number of elements of target tensor) and subsequently only selecting the diagonal elements.

Instead, this change only computes the diagonal elements from the beginning (tensor with just n elements).

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?
• 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 🙃

---

📚 Documentation preview 📚: https://torchmetrics--2346.org.readthedocs.build/en/2346/

[codecov]

Codecov Report

Merging #2346 (ef31bf3) into master (dc115fe) will decrease coverage by 29%.
Report is 1 commits behind head on master.
The diff coverage is 100%.

Additional details and impacted files

@@           Coverage Diff            @@
##           master   #2346     +/-   ##
========================================
- Coverage      69%     40%    -29%     
========================================
  Files         305     305             
  Lines       17149   17149             
========================================
- Hits        11832    6851   -4981     
- Misses       5317   10298   +4981     

[SkafteNicki]

LGTM :)