Add support for fill_value and set_empty_bucket_to in BucketResampler get_sum

[ameraner]

This PR adds support for fill_value and set_empty_bucket_to in BucketResampler get_sum. This is a nice feature, but most importantly it fixes the satpy bucket_sum interface.

Closes pytroll/satpy#2805, closes pytroll/satpy#1746, closes #354

• Tests added
• Tests passed
• Passes git diff origin/main **/*py | flake8 --diff
• Fully documented

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 93.99%. Comparing base (56f4506) to head (05920a8).
Report is 22 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #602      +/-   ##
==========================================
- Coverage   94.01%   93.99%   -0.03%     
==========================================
  Files          92       86       -6     
  Lines       13836    13782      -54     
==========================================
- Hits        13008    12954      -54     
  Misses        828      828              

Flag	Coverage Δ
unittests	93.99% <100.00%> (-0.03%)	⬇️

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

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[coveralls]

coverage: 93.69% (+0.01%) from 93.676%
when pulling 512a265 on ameraner:fix_bucket_sum_for_fillvalue
into db94a87 on pytroll:main.

[coveralls]

coverage: 93.69% (+0.01%) from 93.676%
when pulling 6fc18c6 on ameraner:fix_bucket_sum_for_fillvalue
into db94a87 on pytroll:main.

[ameraner]

I addressed all comments, and most importantly I switched all bucket tests to pytest. I realised there are many ways to do this (like do we need a base test class containing all tests?), so please let me know if you like it like this. I also used fixtures with module scope to optimise it a bit.

I also parametrised the get_sum arguments tests into one test as proposed. It took a while to figure out the combination and simplification of the test flow, but indeed I think it improved the test readability, and made the testing of more corner cases much easier. So thanks for pushing me to do it :)

[coveralls]

coverage: 93.689% (+0.01%) from 93.676%
when pulling f3acfb4 on ameraner:fix_bucket_sum_for_fillvalue
into db94a87 on pytroll:main.

[djhoese]

Small suggestions, but otherwise looks good. I'd really like @pnuu to review this given he's the original author (right?).

[pnuu]

LGTM, just the one nitpick-y change that Dave pointed out.

[djhoese]

@ameraner I'd like to make a release soon. Any idea if you'll have time to finish this PR up soon? Looks like it might just be small things left to change. Also, does this PR also close #364?

[djhoese]

CC @zxdawn about comparisons to #364

[ameraner]

Hi @djhoese , thanks for the review and the ping, indeed it would be great to have this in the next release. I have pushed the modifications as requested.

I believe it should close the PR, yes, as it implements the prosposed functionalities, even with a better granularity using the fill_value and empty_bucket_value arguments.

[coveralls]

coverage: 93.692% (+0.02%) from 93.676%
when pulling 05920a8 on ameraner:fix_bucket_sum_for_fillvalue
into db94a87 on pytroll:main.

[zxdawn]

I agree this can close my old PR because it considers both nan and fill_value. Thanks!