Byte attr support

[hollymandel]

• Closes netCDF4: support byte strings as attribute values #7186
• Adds support for bytes as dataset attributes
• Because h5netcdf does not support arbitrary binaries as attributes, adds a check that the binary is an encoded utf8 free of null characters if this is the engine

[max-sixty]

Thank you!

Could we add a test?

[hollymandel]

I've made the requested changes + cleaned up the commit history

[max-sixty]

Thanks @hollymandel !

I'll merge shortly. Please feel free to add a brief note to whatsnew...

[dcherian]

One way to clean this up would be to rewrite this to

    def test_byte_attrs(self, byte_attrs_dataset) -> None:
        # TODO: consider using self.roundtrip
        with create_tmp_file() as tmp_file:
            self.save(ds, tmp_file)
            with self.open(tmp_file) as ds_out:
                assert ds_out["x"].attrs["null_byte"] == ""
                assert ds_out["x"].attrs["other_bytes"] == other_bytes.decode(
                    errors="replace"
                )

and in the h5netcdf tests we could override

def test_byte_attrs(self, byte_attrs_dataset):
    with pytest.raises(ValueError, ADD_REGEX_HERE):
        super().test_byte_attrs(dataset)

and then elsewhere (conftest.py) create the dataset as a fixture

@pytest.fixture
def byte_attrs_dataset():
null_byte = b"\x00"
    other_bytes = bytes(range(1, 256))
    ds = Dataset({"x": 1}, coords={"x_coord": [1]})
    ds["x"].attrs["null_byte"] = null_byte
    ds["x"].attrs["other_bytes"] = other_bytes
	return ds

[hollymandel]

Awesome, I've made these changes! Really appreciate the feedback.

[dcherian]

Nice work. Thanks!

[max-sixty]

Very nice!