clib.conversion._to_numpy: Add tests for pyarrow.array with date32/date64 dtype

[seisman]

Description of proposed changes

Add tests for pyarrow.array with date64/date32 dtype.

These two types can be converted directly to numpy datetime64 with correct date units, as shown below.

In [1]: from datetime import datetime, date

In [2]: import numpy as np

In [3]: import pyarrow as pa

In [4]: x = pa.array([datetime(2012, 1, 1), date(2013, 1, 1)], type=pa.date32())

In [5]: x
Out[5]:
<pyarrow.lib.Date32Array object at 0x7f7142b70c40>
[
  2012-01-01,
  2013-01-01
]

In [6]: np.ascontiguousarray(x)
Out[6]: array(['2012-01-01', '2013-01-01'], dtype='datetime64[D]')

In [7]: x = pa.array([datetime(2012, 1, 1), date(2013, 1, 1)], type=pa.date64())

In [8]: x
Out[8]:
<pyarrow.lib.Date64Array object at 0x7f726ffb8ca0>
[
  2012-01-01,
  2013-01-01
]

In [9]: np.ascontiguousarray(x)
Out[9]:
array(['2012-01-01T00:00:00.000', '2013-01-01T00:00:00.000'],
      dtype='datetime64[ms]')