Following seems to work at least in some cases:
series.dtype == np.dtype('<M8[ns]')
but it doesn't look pretty and I'm not sure if it works always (with all(?) kinds of Timestamps).
Is there a nicer way for testing whether a column contains Timestamps?