[python-package] Suppress confusing warning in dask module

[wohoef]

This commit prevents UserWarning: Parameter n_jobs will be ignored. from being emitted each time a model is trained with the lightgbm.dask estimators.

The warning is now only shown if a user provides a specific, non-default value (neither None nor -1).

Fixes #6797

[jameslamb]

Thanks for working on this! I support this change.

Please see one suggestion I left. In addition, can you please add tests confirming that the warning is raised and everything otherwise works in these situations?

You can use this as a reference:

LightGBM/tests/python_package_test/test_dask.py

Lines 1191 to 1201 in 915c270

	def test_warns_and_continues_on_unrecognized_tree_learner(cluster):
	with Client(cluster) as client:
	X = da.random.random((1e3, 10))
	y = da.random.random((1e3, 1))
	dask_regressor = lgb.DaskLGBMRegressor(
	client=client, time_out=5, tree_learner="some-nonsense-value", n_estimators=1, num_leaves=2
	)
	with pytest.warns(UserWarning, match="Parameter tree_learner set to some-nonsense-value"):
	dask_regressor = dask_regressor.fit(X, y)
	assert dask_regressor.fitted_

If you're not willing or able to add those tests, just @ me and I'd be happy to push some tests to your branch.

[wohoef]

Thank you for the feedback James. I'll try to see if I can work on writing some tests :)