Add a simple noop test for sklearn API conformance

The "default constructible" (i.e. no required __init__ parameters)
property is the most basic property of a scikit-learn estimator. It is a
prerequisite for all other estimator checks. None of our estimators
adhere to it yet, so this test serves as a todo list for now.

csrank/tests/test_estimators.py

@@ -0,0 +1,49 @@
+"""Check that our estimators adhere to the scikit-learn interface.
+
+https://scikit-learn.org/stable/developers/develop.html#rolling-your-own-estimator
+"""
+
+import pytest
+from sklearn.utils.estimator_checks import check_parameters_default_constructible
+
+from csrank.choicefunction import FETALinearChoiceFunction
+from csrank.discretechoice import FETALinearDiscreteChoiceFunction
+from csrank.objectranking import FETALinearObjectRanker
+
+
+@pytest.mark.parametrize(
+    "Estimator",
+    [
+        # CmpNet,
+        # CmpNetChoiceFunction,
+        # CmpNetDiscreteChoiceFunction,
+        # ExpectedRankRegression,
+        # FATEChoiceFunction,
+        # FATEDiscreteChoiceFunction,
+        # FATELinearChoiceFunction,
+        # FATELinearDiscreteChoiceFunction,
+        # FATELinearObjectRanker,
+        # FATEObjectRanker,
+        # FETAChoiceFunction,
+        # FETADiscreteChoiceFunction,
+        FETALinearChoiceFunction,
+        FETALinearDiscreteChoiceFunction,
+        FETALinearObjectRanker,
+        # FETAObjectRanker,
+        # GeneralizedLinearModel,
+        # GeneralizedNestedLogitModel,
+        # ListNet,
+        # MixedLogitModel,
+        # MultinomialLogitModel,
+        # NestedLogitModel,
+        # PairedCombinatorialLogit,
+        # PairwiseSVMChoiceFunction,
+        # PairwiseSVMDiscreteChoiceFunction,
+        # RankNet,
+        # RankNetChoiceFunction,
+        # RankNetDiscreteChoiceFunction,
+        # RankSVM,
+    ],
+)
+def test_all_estimators(Estimator):
+    check_parameters_default_constructible("default_constructible", Estimator)