test__get_vector_similarity_query_support_true and test__get_vector_similarity_query_support_false for OpenSearchDocumentStore are flaky

[bogdankostic]

Describe the bug
In this test run on the main branch, the tests test__get_vector_similarity_query_support_true and test__get_vector_similarity_query_support_false failed due to an AssertionError. This error didn't occur before.

Error message

E       AssertionError: assert {'script_scor...'knn_score'}}} == {'script_scor...'knn_score'}}}
E         Differing items:
E         {'script_score': {'query': {'match_all': {}}, 'script': {'lang': 'knn', 'params': {'field': 'FooField', 'query_value': [[0.0, 8.4368011761e-313], [1.3498881881773e-311, nan]], 'space_type': 'innerproduct'}, 'source': 'knn_score'}}} != {'script_score': {'query': {'match_all': {}}, 'script': {'lang': 'knn', 'params': {'field': 'FooField', 'query_value': [[0.0, 8.4368011761e-313], [1.3498881881773e-311, nan]], 'space_type': 'innerproduct'}, 'source': 'knn_score'}}}

Additional context
This error most likely is due to the fact that we create a random numpy array here using np.ndarray. As stated in the numpy documentation, this array might contain nan values. And as stated in this section of the numpy documentation, comparing two nan values for equality will always have False as a result.

We should make sure that the randomly initialized numpy array does not contain any nan values, for example by using np.random.random_sample.