Drop PyTorch 1.9 support

[carmocca]

What does this PR do?

See title

Fixes #15346

Does your PR introduce any breaking changes? If yes, please list them.

Drops PyTorch 1.9 support, as expected by our support window

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)
• [n/a] Did you write any new necessary tests? (not for typos and docs)
• [n/a] Did you verify new and existing tests pass locally with your changes?
• Did you list all the breaking changes introduced by this pull request?
• Did you update the CHANGELOG? (not for typos, docs, test updates, or minor internal changes/refactors)

PR review

• 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

cc @carmocca @akihironitta @Borda @justusschock @awaelchli @rohitgr7

[otaj]

PyTorch 1.13 is not yet out officially, so we shouldn't be dropping 1.9 yet

[carmocca]

I know. The release date is supposedly today

[carmocca]

Marking as ready to run tests. Still WIP

[Borda]

@carmocca, how much is missing here?
UPDATE: with #15403 are free to drop support of 1.9 🦦

[carmocca]

This needs to be split into 2:

• Drop 1.9 support
• Add 1.13 testing

The latter is blocked by issues with IPUs and colossalai

[awaelchli]

@carmocca For the latter, is the issue that these libraries don't provide a package with support for 1.13, or is it our implementation that needs updates?
How about we keep the IPU job running on 1.12? And run colossalai tests only on <= 1.12? If Graphcore and Colossal don't release new packages before our next release, we can implement checks that inform the user that these strategies are not yet supported on pytorch >= 1.13.

Since these are optional dependencies for Lightning, I don't think they should block us from providing support and testing of core for the latest pytorch version.

[carmocca]

I've removed all 1.13 related changes so that this can go before #15403.
Note that we need to update IPUs because they use 1.9

[Borda]

Except the TPU docker removal