Fix comparison error for pandas dataframe dtype

[Shelnutt2]

The dtype passed in is likely to be of numpy.dtype when the dtype comes directly from pandas. This means we need to check the type before comparing to the string "ascii".

[shortcut-integration]

This pull request has been linked to Shortcut Story #17286: TypeError: data type 'ascii' not understood.

[nguyenv]

I can confirm this fixes a test that we've seen fail when using NumPy 1.20 (but passes on NumPy 1.22 - latest release). We'll be updating CI so that it tests against multiple versions of NumPy.

(tiledb-clean) vivian@mangonada:~/TileDB-Py$ pip list | grep numpy
numpy            1.20.0
(tiledb-clean) vivian@mangonada:~/TileDB-Py$ git branch --show-current
dev
(tiledb-clean) vivian@mangonada:~/TileDB-Py$ pytest -k "test_sparse_index_dtypes[i8]"
================================================ test session starts ================================================
platform linux -- Python 3.9.12, pytest-7.1.2, pluggy-1.0.0
rootdir: /home/vivian/TileDB-Py, configfile: pyproject.toml, testpaths: tiledb/tests
plugins: hypothesis-6.45.1
collected 450 items / 449 deselected / 1 skipped / 1 selected

tiledb/tests/test_libtiledb.py F                                                                              [100%]

===================================================== FAILURES ======================================================
___________________________________ TestSparseArray.test_sparse_index_dtypes[i8] ____________________________________

self = <tiledb.tests.test_libtiledb.TestSparseArray object at 0x7ff47b0f9340>, dtype = 'i8'

    @pytest.mark.skipif(not has_pandas(), reason="pandas not installed")
    @pytest.mark.parametrize("dtype", INTEGER_DTYPES)
    def test_sparse_index_dtypes(self, dtype):
        path = self.path()
        data = np.arange(0, 3).astype(dtype)

>       schema = schema_from_dict(attrs={"attr": data}, dims={"d0": data})

tiledb/tests/test_libtiledb.py:2248:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
tiledb/util.py:37: in schema_from_dict
    return _sparse_schema_from_dict(attrs, dims)
tiledb/util.py:8: in _sparse_schema_from_dict
    attr_infos = {k: ColumnInfo.from_values(v) for k, v in input_attrs.items()}
tiledb/util.py:8: in <dictcomp>
    attr_infos = {k: ColumnInfo.from_values(v) for k, v in input_attrs.items()}
tiledb/dataframe_.py:109: in from_values
    return cls.from_dtype(array_like.dtype, varlen_types)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

cls = <class 'tiledb.dataframe_.ColumnInfo'>, dtype = dtype('int64'), varlen_types = ()

    @classmethod
    def from_dtype(cls, dtype, varlen_types=()):
        from pandas.api import types as pd_types

>       if dtype == "ascii":
E       TypeError: data type 'ascii' not understood

tiledb/dataframe_.py:115: TypeError
============================================== short test summary info ==============================================
FAILED tiledb/tests/test_libtiledb.py::TestSparseArray::test_sparse_index_dtypes[i8] - TypeError: data type 'ascii...
=================================== 1 failed, 1 skipped, 449 deselected in 0.93s ====================================

(tiledb-clean) vivian@mangonada:~/TileDB-Py$ git branch --show-current
sethshelnutt/sc-17286/typeerror-data-type-ascii-not-understood
(tiledb-clean) vivian@mangonada:~/TileDB-Py$ pip list | grep numpy
numpy            1.20.0
(tiledb-clean) vivian@mangonada:~/TileDB-Py$ git branch --show-current
sethshelnutt/sc-17286/typeerror-data-type-ascii-not-understood
(tiledb-clean) vivian@mangonada:~/TileDB-Py$ pytest -k "test_sparse_index_dtypes[i8]"
================================================ test session starts ================================================
platform linux -- Python 3.9.12, pytest-7.1.2, pluggy-1.0.0
rootdir: /home/vivian/TileDB-Py, configfile: pyproject.toml, testpaths: tiledb/tests
plugins: hypothesis-6.45.1
collected 449 items / 448 deselected / 1 skipped / 1 selected

tiledb/tests/test_libtiledb.py .                                                                              [100%]

=================================== 1 passed, 1 skipped, 448 deselected in 0.80s ====================================

[nguyenv]

[sc-17286]
[sc-16509]

[shortcut-integration]

This pull request has been linked to Shortcut Story #16509: Test failure against numpy 1.20.