[train][tune] Fix LightGBM v2 callbacks for Tune only usage

[liulehui]

Why are these changes needed?

1. in the ray train revamp REP, we decouple the ray train/ray tune dependency.
2. Hence, when using RayTrainReportCallback when reporting metrics or checkpoint: the v2 context api will throw RuntimeError that TrainFnUtils is not found.
3. in this PR, refactor the Callback by inheriting the same base class but using ray.tune.report for tune only and ray.train.report for RayTrainReportCallback based on migration example here to further differentiate these callbacks.
4. ran with testing script: https://gist.github.com/liulehui/58001000fa195c8f000a2992ba3c77e1 and output in the gist comment.

Related issue number

Checks

• I've signed off every commit(by using the -s flag, i.e., git commit -s) in this PR.
• I've run pre-commit jobs to lint the changes in this PR. (pre-commit setup)
• I've included any doc changes needed for https://docs.ray.io/en/master/.
  • I've added any new APIs to the API Reference. For example, if I added a
    method in Tune, I've added it in doc/source/tune/api/ under the
    corresponding .rst file.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
• Testing Strategy
  • Unit tests
  • Release tests
  • This PR is not tested :(

---

Note

Refactors LightGBM callback into a shared base and implements separate Train and Tune callbacks with correct checkpointing/reporting for each API.

• LightGBM callbacks:
  • Refactor: Extract common logic into RayReportCallback with abstract methods for checkpointing and reporting.
  • Train: RayTrainReportCallback now subclasses RayReportCallback, using ray.train.report and rank-aware checkpointing.
  • Tune: Add TuneReportCheckpointCallback subclass using ray.tune.report and tune.Checkpoint (no rank check), replacing prior alias.
  • Minor doc updates: examples and return docs adjusted; references clarified to LightGBM.

^{Written by Cursor Bugbot for commit 8e6abfb. This will update automatically on new commits. Configure here.}

[gemini-code-assist]

Code Review

This pull request refactors the LightGBM callbacks to decouple Ray Tune and Ray Train usage, which is a great improvement. A base RayReportCallback is introduced with specific implementations for Tune and Train. My review focuses on improving the new abstractions and reducing code redundancy. I've pointed out an incorrect method signature in the new abstract base class and suggested removing redundant __init__ methods in the subclasses to make the code cleaner and more maintainable.

[justinvyu]

Thanks!