[dask] Change document to avoid using default import. (#9742)

This aligns dask with pyspark, users need to explicitly call:

```
from xgboost.dask import DaskXGBClassifier
from xgboost import dask as dxgb
```

In future releases, we might stop using the default import and remove the lazy loader.

demo/dask/cpu_survival.py

@@ -9,7 +9,7 @@ import os
 import dask.dataframe as dd
 from dask.distributed import Client, LocalCluster
 
-import xgboost as xgb
+from xgboost import dask as dxgb
 from xgboost.dask import DaskDMatrix
 
 
@@ -48,14 +48,14 @@ def main(client):
         "lambda": 0.01,
         "alpha": 0.02,
     }
-    output = xgb.dask.train(
+    output = dxgb.train(
         client, params, dtrain, num_boost_round=100, evals=[(dtrain, "train")]
     )
     bst = output["booster"]
     history = output["history"]
 
     # you can pass output directly into `predict` too.
-    prediction = xgb.dask.predict(client, bst, dtrain)
+    prediction = dxgb.predict(client, bst, dtrain)
     print("Evaluation history: ", history)
 
     # Uncomment the following line to save the model to the disk

demo/dask/cpu_training.py

@@ -6,7 +6,7 @@ Example of training with Dask on CPU
 from dask import array as da
 from dask.distributed import Client, LocalCluster
 
-import xgboost as xgb
+from xgboost import dask as dxgb
 from xgboost.dask import DaskDMatrix
 
 
@@ -25,7 +25,7 @@ def main(client):
     # distributed version of train returns a dictionary containing the
     # resulting booster and evaluation history obtained from
     # evaluation metrics.
-    output = xgb.dask.train(
+    output = dxgb.train(
         client,
         {"verbosity": 1, "tree_method": "hist"},
         dtrain,
@@ -36,7 +36,7 @@ def main(client):
     history = output["history"]
 
     # you can pass output directly into `predict` too.
-    prediction = xgb.dask.predict(client, bst, dtrain)
+    prediction = dxgb.predict(client, bst, dtrain)
     print("Evaluation history:", history)
     return prediction
 

demo/dask/dask_callbacks.py

@@ -8,6 +8,7 @@ from dask_ml.datasets import make_regression
 from dask_ml.model_selection import train_test_split
 
 import xgboost as xgb
+import xgboost.dask as dxgb
 from xgboost.dask import DaskDMatrix
 
 
@@ -61,7 +62,7 @@ def main(client):
     dtrain = DaskDMatrix(client, X_train, y_train)
     dtest = DaskDMatrix(client, X_test, y_test)
 
-    output = xgb.dask.train(
+    output = dxgb.train(
         client,
         {
             "verbosity": 1,

demo/dask/gpu_training.py

@@ -8,7 +8,6 @@ from dask import dataframe as dd
 from dask.distributed import Client
 from dask_cuda import LocalCUDACluster
 
-import xgboost as xgb
 from xgboost import dask as dxgb
 from xgboost.dask import DaskDMatrix
 
@@ -21,7 +20,7 @@ def using_dask_matrix(client: Client, X: da.Array, y: da.Array) -> da.Array:
     # Use train method from xgboost.dask instead of xgboost.  This distributed version
     # of train returns a dictionary containing the resulting booster and evaluation
     # history obtained from evaluation metrics.
-    output = xgb.dask.train(
+    output = dxgb.train(
         client,
         {
             "verbosity": 2,
@@ -37,7 +36,7 @@ def using_dask_matrix(client: Client, X: da.Array, y: da.Array) -> da.Array:
     history = output["history"]
 
     # you can pass output directly into `predict` too.
-    prediction = xgb.dask.predict(client, bst, dtrain)
+    prediction = dxgb.predict(client, bst, dtrain)
     print("Evaluation history:", history)
     return prediction
 
@@ -56,14 +55,14 @@ def using_quantile_device_dmatrix(client: Client, X: da.Array, y: da.Array) -> d
     # be used for anything else other than training unless a reference is specified. See
     # the `ref` argument of `DaskQuantileDMatrix`.
     dtrain = dxgb.DaskQuantileDMatrix(client, X, y)
-    output = xgb.dask.train(
+    output = dxgb.train(
         client,
         {"verbosity": 2, "tree_method": "hist", "device": "cuda"},
         dtrain,
         num_boost_round=4,
     )
 
-    prediction = xgb.dask.predict(client, output, X)
+    prediction = dxgb.predict(client, output, X)
     return prediction
 
 

demo/dask/sklearn_cpu_training.py

@@ -5,7 +5,7 @@ Use scikit-learn regressor interface with CPU histogram tree method
 from dask import array as da
 from dask.distributed import Client, LocalCluster
 
-import xgboost
+from xgboost import dask as dxgb
 
 
 def main(client):
@@ -16,7 +16,7 @@ def main(client):
     X = da.random.random((m, n), partition_size)
     y = da.random.random(m, partition_size)
 
-    regressor = xgboost.dask.DaskXGBRegressor(verbosity=1, n_estimators=2)
+    regressor = dxgb.DaskXGBRegressor(verbosity=1, n_estimators=2)
     regressor.set_params(tree_method="hist")
     # assigning client here is optional
     regressor.client = client

demo/dask/sklearn_gpu_training.py

@@ -9,7 +9,7 @@ from dask.distributed import Client
 # It's recommended to use dask_cuda for GPU assignment
 from dask_cuda import LocalCUDACluster
 
-import xgboost
+from xgboost import dask as dxgb
 
 
 def main(client):
@@ -20,7 +20,7 @@ def main(client):
     X = da.random.random((m, n), partition_size)
     y = da.random.random(m, partition_size)
 
-    regressor = xgboost.dask.DaskXGBRegressor(verbosity=1)
+    regressor = dxgb.DaskXGBRegressor(verbosity=1)
     # set the device to CUDA
     regressor.set_params(tree_method="hist", device="cuda")
     # assigning client here is optional