bug: Ml nodata

[mmann1123]

What is this PR changing?

• fixes install issue with pip [ml]
• avoids predicting empty rasters Resolves Ml no data #270
• throws error if stack_dim is time. Resolves polygon_to_array returns shorter y dimension than source rasters #311

Checklist

• [ x] Remember to add a semantic tag to the commit name

[jgrss]

Is there any way we can make this scikit-learn>=1.2.0?

[mmann1123]

Unfortunately we need to pin that for the moment. It's bc sklearm_xarray is so outdated

[jgrss]

Do you know what needs updated in sklearn-xarray? We are installing a forked version so we can modify it.

I'm a bit hesitant to enforce such a hard dependency pin.

[mmann1123]

At the moment I don't think we have much of an option. My plan is to take a look at this in about 2 weeks when I get done with classes. On a positive note, I think this only comes into play with the machine learning add-ons. So it should n't affect very many users for the moment.

[jgrss]

I'm slightly confused, because I thought we recently modified this. In main it is

    def test_union_values(self):
        filenames = [l8_224077_20200518_B2, l8_224078_20200518_B2]
        with gw.open(
            filenames,
            band_names=['blue'],
            mosaic=True,
            bounds_by='union',
        ) as src:
            values = src.values[
                0,
                src.gw.nrows // 2,
                src.gw.ncols // 2 : src.gw.ncols // 2 + 10,
            ]
            self.assertTrue(
                np.allclose(
                    values,
                    np.array(
                        [
                            7524,
                            7538,
                            7573,
                            7625,
                            7683,
                            7661,
                            7643,
                            7773,
                            7697,
                            7566,
                        ]
                    ),
                )
            )

[mmann1123]

Sorry I think I started the branch right before you pushed the changes. Sorry about that.

[jgrss]

Ah, okay, no worries. I can pull in the updated tests from main.

[mmann1123]

I don't have the ability to merge here, assuming you will bundle it with the others