Reintroduce test sampling with MVN score

Author: han-ol

tests/test_approximators/test_point_approximators/test_sample.py

@@ -1,15 +1,12 @@
 import keras
 import numpy as np
 from bayesflow.scores import ParametricDistributionScore
-from tests.utils import check_combination_simulator_adapter, check_approximator_multivariate_normal_score
+from tests.utils import check_combination_simulator_adapter
 
 
 def test_approximator_sample(point_approximator, simulator, batch_size, num_samples, adapter):
     check_combination_simulator_adapter(simulator, adapter)
 
-    # as long as MultivariateNormalScore is unstable, skip test
-    check_approximator_multivariate_normal_score(point_approximator)
-
     data = simulator.sample((batch_size,))
 
     batch = adapter(data)