Fixing flaky test (#780)

Summary:
Running some tests by removing seeds revealed that tests such as `test_single_shapley_int_batch_scalar_float`, `test_single_shapley_int_batch_scalar_tensor_0d`, `test_single_shapley_int_batch_scalar_tensor_1d`, and `test_single_shapley_int_batch_scalar_tensor_int` are flaky. They almost fail 40-50% of the times.

To fix these tests, I updated the `n_samples` to 2500 to fix the flakiness. The tests now pass consistently. This only affects the execution time of the test by ~1s.

Please let me know if this change seems reasonable. Similar to #775.

Pull Request resolved: #780

Reviewed By: NarineK, Reubend

Differential Revision: D31587778

Pulled By: vivekmig

fbshipit-source-id: 843adda41662e881a6bf81ebc4354d6f304e83ce

Author: crawlingcub

tests/attr/test_shapley.py

@@ -349,6 +349,7 @@ def _single_int_input_multi_sample_batch_scalar_shapley_assert(
             feature_mask=mask,
             perturbations_per_eval=(1,),
             target=None,
+            n_samples=2500,
         )
 
     def _multi_input_batch_scalar_shapley_assert(self, func: Callable) -> None: