Fix declustering output

[dnerini]

Closes #194

The issue described in #194 was caused by a (seemingly unnecessary) squeezing of the output (see 810d569).

Additionally:

• Refactor tests for utils module
• Add tests for utils.cleansing submodule
• Test for negative outliers, too, in the cleansing.detect_outliers univariate case
• Fix wrong shape of output in cleansing.detect_outliers in case of LinAlgErr
• Simplify syntax for the computation of uniques pairs of coordinates in cleansing.decluster (c1521a5)

[codecov[bot]]

Codecov Report

Merging #195 (5250b38) into master (e2c3484) will increase coverage by 0.44%. The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master     #195      +/-   ##
==========================================
+ Coverage   75.90%   76.34%   +0.44%     
==========================================
  Files         126      130       +4     
  Lines        9561     9686     +125     
==========================================
+ Hits         7257     7395     +138     
+ Misses       2304     2291      -13     

Flag	Coverage Δ
unit_tests	76.34% <100.00%> (+0.44%)	:arrow_up:

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

Impacted Files	Coverage Δ
pysteps/motion/lucaskanade.py	93.42% <100.00%> (ø)
pysteps/tests/test_utils_arrays.py	100.00% <100.00%> (ø)
pysteps/tests/test_utils_cleansing.py	100.00% <100.00%> (ø)
pysteps/tests/test_utils_conversion.py	100.00% <100.00%> (ø)
pysteps/tests/test_utils_dimension.py	100.00% <100.00%> (ø)
pysteps/tests/test_utils_transformation.py	100.00% <100.00%> (ø)
pysteps/utils/cleansing.py	81.81% <100.00%> (+13.45%)	:arrow_up:
... and 3 more

---

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 e2c3484...5250b38. Read the comment docs.