The previous implementation was clearly wrong (the isinstance check would raise TypeError as the second argument would be a bool), but the tests didn't catch that because the bug led to _is_tensorflow_array returning False, then _unpack_to_numpy returning the original input, and then assert_array_equal implicitly convertingresult by calling__array__ on it. Fix the test by explicitly checking thatresult is indeed a numpy array, and also fix _is_tensorflow_array with more restrictive exception catching (also applied to _is_torch_array, _is_jax_array, and _is_pandas_dataframe, while we're at it).

PR summary

PR checklist

• "closes #0000" is in the body of the PR description tolink the related issue
• new and changed code istested
• Plotting related features are demonstrated in anexample
• New Features andAPI Changes are noted with adirective and release note
• Documentation complies withgeneral anddocstring guidelines