[dask] warn if attempting to use tree_learner other than data parallel (

#3848)

* [dask] warn if attempting to use tree_learn other than data parallel

* add param

python-package/lightgbm/dask.py

@@ -244,6 +244,7 @@ def _train(client, data, label, params, model_factory, sample_weight=None, group
     for tree_learner_param in _ConfigAliases.get('tree_learner'):
         tree_learner = params.get(tree_learner_param)
         if tree_learner is not None:
+            params['tree_learner'] = tree_learner
             break
 
     allowed_tree_learners = {
@@ -261,6 +262,11 @@ def _train(client, data, label, params, model_factory, sample_weight=None, group
         _log_warning('Parameter tree_learner set to %s, which is not allowed. Using "data" as default' % tree_learner)
         params['tree_learner'] = 'data'
 
+    if params['tree_learner'] not in {'data', 'data_parallel'}:
+        _log_warning(
+            'Support for tree_learner %s in lightgbm.dask is experimental and may break in a future release. Use "data" for a stable, well-tested interface.' % params['tree_learner']
+        )
+
     local_listen_port = 12400
     for port_param in _ConfigAliases.get('local_listen_port'):
         val = params.get(port_param)