Use dask to reduce memory consumption of extract_levels for masked data

[valeriupredoi]

Before you start, please read our contribution guidelines.

Tasks

• Create an issue to discuss what you are going to do, if you haven't done so already (and add the link at the bottom)
• This pull request has a descriptive title that can be used in a changelog
• Add unit tests
• Public functions should have a numpy-style docstring so they appear properly in the API documentation. For all other functions a one line docstring is sufficient.
• If writing a new/modified preprocessor function, please update the documentation
• Circle/CI tests pass. Status can be seen below your pull request. If the tests are failing, click the link to find out why.
• Codacy code quality checks pass. Status can be seen below your pull request. If there is an error, click the link to find out why. If you suspect Codacy may be wrong, please ask by commenting.

If you need help with any of the tasks above, please do not hesitate to ask by commenting in the issue or pull request.

---

partially addressing #775

Mostly appreciated by ocean people, I noticed a memory consumption reduction by about 30% for level selection.

[valeriupredoi]

can one of you @bouweandela @jvegasbsc @schlunma pls review this, it'd be useful to have it in the release 👍 🍺

[bouweandela]

Could you add a test like this one, but with lazy input and output data?

ESMValCore/tests/integration/preprocessor/_regrid/test_extract_levels.py

Lines 48 to 57 in a6d4bbc

	def test_interpolation__linear(self):
	levels = [0.5, 1.5]
	scheme = 'linear'
	result = extract_levels(self.cube, levels, scheme)
	expected = np.array([[[[2., 3.], [4., 5.]], [[6., 7.], [8., 9.]]],
	[[[14., 15.], [16., 17.]], [[18., 19.],
	[20., 21.]]]])
	self.assert_array_equal(result.data, expected)
	self.shape[self.z_dim] = len(levels)
	self.assertEqual(result.shape, tuple(self.shape))

[bouweandela]

Oh wait, that won't work, because this doesn't make the process completely lazy, right?

[valeriupredoi]

Oh wait, that won't work, because this doesn't make the process completely lazy, right?

yer, it's a half-mule process due to vinterp not returning a lazy object

[valeriupredoi]

@bouweandela 👍 or 👎

[bouweandela]

@bouweandela +1 or -1

Please see #776 (comment)

[bouweandela]

To make a cube with realized data lazy, you can do

cube.data = cube.lazy_data()

or

cube.data = da.asarray(cube.data, chunks=(1, 2, 3, 4))

if you want more control over how the lazy array is created. There is no need to first save to file.

[valeriupredoi]

cool! I didn't know that, but I wanted to replicate the actual conditions the function runs in anyway