[python-package][dask] avoid warning for default n_jobs values

[shiyu1994]

Summary

• avoid warning in lightgbm.dask._train() when a num_threads alias is present with default sklearn-constructor values (None or -1)
• keep warning behavior unchanged for other explicit num_threads / num_machines values that are ignored
• add a Dask test covering both default n_jobs alias values (-1, None)

Testing

• python -m compileall -q python-package/lightgbm/dask.py tests/python_package_test/test_dask.py
• targeted test_dask.py run could not be executed here because the available LightGBM test environment lacks dask / distributed and package install is blocked by network proxy

Fixes #6797

[jameslamb]

There is already an open PR attempting to close this, authored by (as far as I can tell) a human who put effort into it: #6987

I'd prefer to take that PR over this LLM-generated one.

It has been 8 months since they last interacted there though. I just asked them to respond. If we don't hear back soon, I'd be ok with closing that PR and pursuing this one. If we do that, please do see my notes about the tests.

[jameslamb]

This does not test the behavior described in #6797

1. doesn't actually check the log output to confirm a warning isn't raised
2. passes parameter n_jobs which is interesting, but the issue says the warning is raised even when not passing any multithreading related parameters

If we move forward with this PR, this needs to be updated to confirm that the issue is actually fixed.