[MNT] speed up clustering dunder test (#4982)

This PR speeds up a costly test for the clustering estimator dunders, by
using a much smaller dataset.
The test time was >1 min and should now be in the range of seconds at
most.
Related to: #2890

Also removes the dependency on `numba` by using a clusterer that depends
only on `sklearn`.

Test coverage does not change, as the requirements are only that the
estimator is a clusterer, and produces a non-trivial clustering (ensured
by the choice of parameters and random seeds).

sktime/clustering/compose/tests/test_pipeline.py

@@ -4,34 +4,28 @@
 __author__ = ["fkiraly"]
 __all__ = []
 
-import pytest
 from sklearn.preprocessing import StandardScaler
 
 from sktime.clustering.compose import ClustererPipeline
-from sktime.clustering.k_means import TimeSeriesKMeans
-from sktime.clustering.kernel_k_means import TimeSeriesKernelKMeans
+from sktime.clustering.dbscan import TimeSeriesDBSCAN
+from sktime.dists_kernels import FlatDist
 from sktime.transformations.panel.padder import PaddingTransformer
 from sktime.transformations.series.exponent import ExponentTransformer
 from sktime.transformations.series.impute import Imputer
 from sktime.utils._testing.estimator_checks import _assert_array_almost_equal
 from sktime.utils._testing.panel import _make_panel_X
-from sktime.utils.validation._dependencies import _check_estimator_deps
 
 
-@pytest.mark.skipif(
-    not _check_estimator_deps(TimeSeriesKernelKMeans, severity="none"),
-    reason="skip test if required soft dependencies not available",
-)
 def test_dunder_mul():
     """Test the mul dunder method."""
     RAND_SEED = 42
-    X = _make_panel_X(n_instances=10, n_timepoints=20, random_state=RAND_SEED)
-    X_test = _make_panel_X(n_instances=5, n_timepoints=20, random_state=RAND_SEED)
+    X = _make_panel_X(n_instances=10, n_timepoints=12, random_state=RAND_SEED)
+    X_test = X
 
     t1 = ExponentTransformer(power=4)
     t2 = ExponentTransformer(power=0.25)
 
-    c = TimeSeriesKernelKMeans(random_state=RAND_SEED)
+    c = TimeSeriesDBSCAN(FlatDist.create_test_instance(), eps=4, min_samples=1)
     t12c_1 = t1 * (t2 * c)
     t12c_2 = (t1 * t2) * c
     t12c_3 = t1 * t2 * c
@@ -47,19 +41,15 @@ def test_dunder_mul():
     _assert_array_almost_equal(y_pred, t12c_3.fit(X).predict(X_test))
 
 
-@pytest.mark.skipif(
-    not _check_estimator_deps(TimeSeriesKMeans, severity="none"),
-    reason="skip test if required soft dependencies not available",
-)
 def test_mul_sklearn_autoadapt():
     """Test auto-adapter for sklearn in mul."""
     RAND_SEED = 42
-    X = _make_panel_X(n_instances=10, n_timepoints=20, random_state=RAND_SEED)
-    X_test = _make_panel_X(n_instances=10, n_timepoints=20, random_state=RAND_SEED)
+    X = _make_panel_X(n_instances=10, n_timepoints=12, random_state=RAND_SEED)
+    X_test = X
 
     t1 = ExponentTransformer(power=2)
     t2 = StandardScaler()
-    c = TimeSeriesKMeans(random_state=RAND_SEED)
+    c = TimeSeriesDBSCAN(FlatDist.create_test_instance(), eps=4, min_samples=1)
 
     t12c_1 = t1 * (t2 * c)
     t12c_2 = (t1 * t2) * c
@@ -75,13 +65,9 @@ def test_mul_sklearn_autoadapt():
     _assert_array_almost_equal(y_pred, t12c_3.fit(X).predict(X_test))
 
 
-@pytest.mark.skipif(
-    not _check_estimator_deps(TimeSeriesKMeans, severity="none"),
-    reason="skip test if required soft dependencies not available",
-)
 def test_missing_unequal_tag_inference():
     """Test that ClustererPipeline infers missing/unequal tags correctly."""
-    c = TimeSeriesKMeans()
+    c = TimeSeriesDBSCAN(FlatDist.create_test_instance())
     c1 = ExponentTransformer() * PaddingTransformer() * ExponentTransformer() * c
     c2 = ExponentTransformer() * ExponentTransformer() * c
     c3 = Imputer() * ExponentTransformer() * c