Adds open_datatree and load_datatree to the tutorial module

[eni-awowale]

Adds open_datatree and load_datatree to the tutorial module

• Closes #xxxx
• Tests added
• User visible changes (including notable bug fixes) are documented in whats-new.rst
• New functions/methods are listed in api.rst

[keewis]

I've got two comments. Other than that, we'll probably have to refactor the way we call pooch since we now basically duplicated the code of open_dataset (doesn't have to be in this PR, though).

[keewis]

it's probably better to just hard-code the dataset name into the test, there's no point in parametrizing this (to be clear, this part of the test suite is pretty old):

Suggested change

	ds = tutorial.open_dataset(testfile, cache_dir=cache_dir).load()
	ds = tutorial.open_dataset("tiny", cache_dir=cache_dir).load()

[keewis]

do the hdf5 file work with both netcdf4 and h5netcdf? Otherwise we might need to specialize, like we do with grib

[eni-awowale]

Yes, imerghh_730.HDF5 and imerghh_830.HDF5, works for both engines. I think if we wanted to add hdf5 files without named dimensions we would want to specify the engine as h5netcdf.

EDIT:
Since pydata/xarray-data#32 was merged, we do have to explicitly add the extension, e.g. xr.tutorial.open_datatree('imerghh_830.hdf5'), otherwise it defaults to .nc

[eni-awowale]

Updated this to use the xarray.testing module's assert_identical because xarray.tests didn't support DataTree objects.

[eni-awowale]

FYI these checks look like they are failing in main 😬

[keewis]

yep, that's a change to array-api-strict. If you want to fix them within the next four hours, I think it should be sufficient to cast the condition of duck_array_ops.where to a bool dtype (otherwise I'll send in a PR myself).

[eni-awowale]

Sure! Is there a issue for this yet? I can get a PR started and you can jump in when you're free.

[keewis]

there's #10084, but nothing else. I think you can just open the PR