CI: Change flaky to_excel test to use xlsxwriter

[mroeschke]

test_write_fspath_all has been flaky and failing when comparing the binary xml for to_excel with openpyxl e.g.

_ TestCommonIOCapabilities.test_write_fspath_all[to_excel-writer_kwargs1-openpyxl] _
[gw2] darwin -- Python 3.9.13 /Users/runner/micromamba/envs/test/bin/python3.9

self = <pandas.tests.io.test_common.TestCommonIOCapabilities object at 0x157198af0>
writer_name = 'to_excel', writer_kwargs = {'engine': 'openpyxl'}
module = 'openpyxl'

    @pytest.mark.filterwarnings("ignore:In future versions `DataFrame.to_latex`")
    @pytest.mark.parametrize(
        "writer_name, writer_kwargs, module",
        [
            ("to_csv", {}, "os"),
            ("to_excel", {"engine": "openpyxl"}, "openpyxl"),
            ("to_feather", {}, "pyarrow"),
            ("to_html", {}, "os"),
            ("to_json", {}, "os"),
            ("to_latex", {}, "os"),
            ("to_pickle", {}, "os"),
            ("to_stata", {"time_stamp": pd.to_datetime("2019-01-01 00:00")}, "os"),
        ],
    )
    def test_write_fspath_all(self, writer_name, writer_kwargs, module):
        p1 = tm.ensure_clean("string")
        p2 = tm.ensure_clean("fspath")
        df = pd.DataFrame({"A": [1, 2]})
    
        with p1 as string, p2 as fspath:
            pytest.importorskip(module)
            mypath = CustomFSPath(fspath)
            writer = getattr(df, writer_name)
    
            writer(string, **writer_kwargs)
            with open(string, "rb") as f:
                expected = f.read()
    
            writer(mypath, **writer_kwargs)
            with open(fspath, "rb") as f:
                result = f.read()
    
>           assert result == expected
E           assert b'PK\x03\x04\...0\x00\x00\x00' == b'PK\x03\x04\...0\x00\x00\x00'
E             At index 10 diff: b'\x13' != b'\x12'
E             Full diff:
E               (
E             -  b'PK\x03\x04\x14\x00\x00\x00\x08\x00\x12xbU\x07AMb\x81\x00\x00\x00\xb1\x00'
E             ?                                         ^
E             +  b'PK\x03\x04\x14\x00\x00\x00\x08\x00\x13xbU\x07AMb\x81\x00\x00\x00\xb1\x00'
E             ?           

This was changed from xlwt to openpyxl in #49296, seeing if xlsxwriter is more stable.

[rhshadrach]

Guessing this should now pass with openpyxl as well as xlsxwriter now that we aren't comparing all file contents. Would it make sense to still test openpyxl, or is that redundant?

[mroeschke]

I think testing both openpyxl and xlsxwriter would be redundant since it seem the main point of the test is to test a CustomFSPath

I changed it back to testing openpyxl.

[rhshadrach]

lgtm

[rhshadrach]

Thanks @mroeschke