[skip ci] Allow ModelCheckpoint in Trainer callbacks list

[awaelchli]

Allows passing ModelCheckpoint directly to the callbacks list:

Trainer(callbacks=[ModelCheckpoint(...)])

This was not possible before because save_function was not set. We now set it in the callback directly.

The following cases need to be considered:

• Trainer(checkpoint_callback=True, callbacks=[ModelCheckpoint()])
• Trainer(checkpoint_callback=False, callbacks=[ModelCheckpoint()])
• Trainer(checkpoint_callback=ModelCheckpoint(), callbacks=[ModelCheckpoint()])

This PR ignores the checkpoint_callback setting if one is passed to callbacks. Does this make sense @PyTorchLightning/core-contributors ?

TODO:

• needs a test
• need to define trainer.checkpoint_callback, we probably need to limit it to 1 checkpoint callback for now?

[codecov]

Codecov Report

Merging #3990 into master will decrease coverage by 47%.
The diff coverage is 82%.

@@           Coverage Diff            @@
##           master   #3990     +/-   ##
========================================
- Coverage      89%     42%    -47%     
========================================
  Files         212     212             
  Lines       15562   15702    +140     
========================================
- Hits        13867    6623   -7244     
- Misses       1695    9079   +7384     

[awaelchli]

The tests currently fail as they reveal another bug, which I will fix here: #4027

[stale]

This pull request has been automatically marked as stale because it has not had recent activity. It will be closed in 7 days if no further activity occurs. If you need further help see our docs: https://pytorch-lightning.readthedocs.io/en/latest/CONTRIBUTING.html#pull-request or ask the assistance of a core contributor here or on Slack. Thank you for your contributions.