Replace stacking gradient search with resample_blocks variant

[mraspaud]

This PR replaces the stacking gradient search with resample blocks for the swath to area case.
It also adds support for area to swath in the gradient search.

• Closes Gradient search resampling swath data gives transposed results #620
• Tests added
• Tests passed
• Fully documented

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 93.96%. Comparing base (c4ce8ee) to head (4dd2948).
Report is 12 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #626      +/-   ##
==========================================
- Coverage   93.97%   93.96%   -0.01%     
==========================================
  Files          86       86              
  Lines       13848    13538     -310     
==========================================
- Hits        13013    12721     -292     
+ Misses        835      817      -18     

Flag	Coverage Δ
unittests	93.96% <100.00%> (-0.01%)	⬇️

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: 93.66% (-0.03%) from 93.686%
when pulling 4dd2948 on mraspaud:feature-remove-stacking-gradient
into 8da073e on pytroll:main.

[pnuu]

Couple of initial comments. I'll do some real-life testing next.

I'm surprised you could get rid of such a huge chunk of code!

[mraspaud]

For reference, main seems to have striping problems:

This branch doesn't:

[mraspaud]

Also, preliminary test, resampling that metop scene to euron1 takes 5GB on main vs 3GB with this PR

[pnuu]

In my tests the memory usage stayed the same, but the PR is much faster, processing time went from 57.6 s to 49.1 s.

main

[pnuu]

Third try, maybe now the correct graph...

PR

[pnuu]

LGTM.

Thanks for a) fixing #620 and b) getting rid of huge amounts of code!