fix self.device access in DataParallel

[awaelchli]

What does this PR do?

DataParallel does not maintain state during the forward on the replicas. Therefore, self.device in e.g. training_step points to the root device no matter on which gpu.
We can fix this problem by taking the device from the input tensors, these are guaranteed to be on the right device.

The only exception is when the model doesn't get a batch with tensors in it. This is rare, and the best we can do is warn the user that self.device will not work.

Fixes #6413
Fixes #6563

Before submitting

• Was this discussed/approved via a GitHub issue? (not for typos and docs)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure your PR does only one thing, instead of bundling different changes together?
• Did you make sure to update the documentation with your changes? (if necessary)
• Did you write any new necessary tests? (not for typos and docs)
• Did you verify new and existing tests pass locally with your changes?
• Did you update the CHANGELOG? (not for typos, docs, test updates, or internal minor changes/refactorings)

PR review

Anyone in the community is free to review the PR once the tests have passed.
Before you start reviewing make sure you have read Review guidelines. In short, see the following bullet-list:

• Is this pull request ready for review? (if not, please submit in draft mode)
• Check that all items from Before submitting are resolved
• Make sure the title is self-explanatory and the description concisely explains the PR
• Add labels and milestones (and optionally projects) to the PR so it can be classified

Did you have fun?

Make sure you had fun coding 🙃

[codecov]

Codecov Report

Merging #6414 (5bd3db0) into master (b46cc55) will decrease coverage by 5%.
The diff coverage is 80%.

@@           Coverage Diff           @@
##           master   #6414    +/-   ##
=======================================
- Coverage      92%     87%    -5%     
=======================================
  Files         194     194            
  Lines       12355   12365    +10     
=======================================
- Hits        11339   10696   -643     
- Misses       1016    1669   +653     

[tchaton]

Really Neat !