Update SEVIRI ICARE reader to properly use dask.

[simonrp84]

The iCARE/SEVIRI reader was not daskified, data was being stored as numpy arrays. This raised an error when trying to create a natural_color image as the Rayleigh correction code was expecting dask arrays.

This PR updates the reader so that the data is stored as a dask array.

[mraspaud]

This is a reboot of #1758

[mraspaud]

Looks good, anything you can add to the tests to make sure the same error doesn't reappear in the future?

[codecov]

Codecov Report

Merging #2103 (0d1fb5e) into main (56eb21b) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##             main    #2103   +/-   ##
=======================================
  Coverage   93.92%   93.92%           
=======================================
  Files         283      283           
  Lines       42715    42722    +7     
=======================================
+ Hits        40120    40127    +7     
  Misses       2595     2595           

Flag	Coverage Δ
behaviourtests	4.70% <0.00%> (-0.01%)	⬇️
unittests	94.48% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
satpy/readers/seviri_l1b_icare.py	90.19% <100.00%> (ø)
satpy/tests/reader_tests/test_seviri_l1b_icare.py	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 56eb21b...0d1fb5e. Read the comment docs.

[pnuu]

The only thing I can think of adding to tests would be the use of satpy.tests.utils import CustomScheduler and reading some simulated data and checking there are no computes. Not sure if that's feasible or not. See https://github.com/pytroll/satpy/blob/main/satpy/tests/test_composites.py#L1305-L1308 for an example of the usage.

[pnuu]

LGTM

[coveralls]

Coverage increased (+0.0009%) to 94.422% when pulling 0d1fb5e on simonrp84:icare_daskify into 56eb21b on pytroll:main.

[simonrp84]

Would something like this be appropriate?

    def test_nocompute(self):
        """Test that dask does not compute anything in the reader itself."""
        from satpy.tests.utils import CustomScheduler
        import dask
        with dask.config.set(scheduler=CustomScheduler(max_computes=0)):
            r = load_reader(self.reader_configs)
            loadables = r.select_files_from_pathnames([
                'GEO_L1B-MSG1_2004-12-29T12-15-00_G_VIS08_V1-04.hdf'
            ])
            r.create_filehandlers(loadables)
            r.load(['VIS008'])

[mraspaud]

That looks about right! Just test this before and after your changes to make sure it works

[simonrp84]

Ok, here we go, dask compute test added.

In adding this test I also discovered that the original tests were using pure numpy arrays rather than dask arrays - that's fixed now too.

With this new test, the pre-PR reader code fails, but the post-PR code passes.

[sfinkens]

LGTM, just one question.

[mraspaud]

LGTM! I'll just subscribe to @sfinkens 's question saw now it was answered already

[mraspaud]

pre-commit.ci autofix