Fix int dtype test

[olebole]

During the update of the Debian package for glue I observed one failure when the test runs on a 32-bit machine:

________________________________ test_dtype_int ________________________________
    def test_dtype_int():
        data = b'# a, b\n1, 1 \n2, 2 \n3, 3'
        with make_file(data, '.csv') as fname:
            d = df.load_data(fname)
>       assert d['a'].dtype == int
E       AssertionError: assert dtype('int64') == int
E        +  where dtype('int64') = array([1, 2, 3], dtype=int64).dtype
glue/core/data_factories/tests/test_data_factories.py:125: AssertionError

This patch fixes this by explicitly using np.int64 instead of int.

[dhomeier]

Thanks for finding this! I suspect this has sneaked in with Numpy 2; just wondering if this still passes with Numpy 1.x on 32bit, but we don't have either in our test matrix.