Fix stateful dataloader state not restored on resume after estimated_stepping_batches

[discobot]

What does this PR do?

Fixes #20550.

Stateful dataloader states saved in a checkpoint are silently dropped on resume when
configure_optimizers accesses trainer.estimated_stepping_batches. Building on the analysis in
the issue: the first setup_data() runs during strategy.setup(), before
restore_training_state(), so when _FitLoop.on_load_checkpoint later stashes the checkpoint's
dataloader states into _combined_loader_states_to_load, the second setup_data() early-returns
(_combined_loader already exists) and _load_combined_loader_states() never runs.

This PR loads the pending states in that early-return branch and re-creates the data-fetcher
iterator. The re-iteration matters: the existing iterator was created before the restore, so
loaders that snapshot their position at iter() time (torchdata StatefulDataLoader-style) — or
a _PrefetchDataFetcher that already prefetched a batch — would otherwise keep stale pre-restore
state. A states-only variant passes the issue's repro but fails an eager-iterator one; loading
then re-iterating matches the order of the normal resume path. The new branch is guarded on
restarting plus pending states, so it is a no-op everywhere else.

Testing: parametrized test_fit_loop_save_and_restore_dataloaders with a flag that accesses
estimated_stepping_batches in configure_optimizers — the stateful variants fail on master and
pass with this change. tests_pytorch/loops/,
trainer/properties/test_estimated_stepping_batches.py, and
utilities/test_combined_loader.py pass locally (CPU). CHANGELOG updated.

Before submitting

• Was this discussed/agreed 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 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

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

Reviewer checklist

• 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