Fix(table_diff): Properly handle null check for array types in data sample

[erindru]

Prior to this, showing the row_diff sample on a table where Pandas had coerced a value to an ndarray would return the following error:

[ValueError: The truth value of an array with more than one element is ambiguous.

The cause is that the current code assumed pd.isnull() would always return a bool, when actually it returns an array[bool] if it gets passed an array.

This PR updates the code to handle arrays and say "arrays are considered null (python None) if all their items are null"