[Python] Recognize pandas arrays in the array() constructor for the from_pandas argument

The `from_pandas` argument in the `array(..)` constructor determines whether NaN-like values are converted to null, or preserved as NaN. If not specified by the user, we will set the keyword automatically when detecting the input as a "pandas object". Currently we only check for Series and Index, but we should also check for ExtensionArrays:

```
>>> pa.array([1.2, np.nan])

[
1.2,
nan
]
>>> pa.array([1.2, np.nan], from_pandas=True)

[
1.2,
null
]
>>> pa.array(pd.Series([1.2, np.nan]))

[
1.2,
null
]
>>> pa.array(pd.array([1.2, np.nan], dtype=float))

[
1.2,
nan
]
```

Start at the Python array() constructor and the logic that detects pandas objects for the from_pandas argument. Reproduce the examples with a pandas ExtensionArray, then verify that automatic detection converts NaN-like values to null while preserving explicit behavior for other inputs.

Written by the indexing model from the issue text.