WIP: Better parallel implementation of linear regression

[jbusecke]

The previous implementation of xr_linregress, required to chunk data in one piece in time, making it wildly unpractical for e.g. long climate simulations (which are typically chunked in time).

This PR implements the basic methodology behind scipy.stats.linregress in mostly xarray.
Extensive testing makes sure that the results are exactly as if scipy.stats.linregress was applied along single vectors along the dimension dim.

I am fast tracking this PR without much documentation, since I need it for a project and also this might have to be refactored, once xarray implements part of this functionality (see e.g. here).

• Tests

  • Pass locally
  • CI

• Docs

  • Minimal example in notebook
  • Entry in whats new (This is a breaking change)

[codecov]

Codecov Report

Merging #62 into master will decrease coverage by 0.18%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master      #62      +/-   ##
==========================================
- Coverage   48.23%   48.05%   -0.19%     
==========================================
  Files          13       13              
  Lines        1391     1388       -3     
  Branches      324      319       -5     
==========================================
- Hits          671      667       -4     
  Misses        665      665              
- Partials       55       56       +1     

Impacted Files	Coverage Δ
xarrayutils/utils.py	39.50% <100.00%> (-0.64%)	⬇️
xarrayutils/_version.py	36.91% <0.00%> (-0.36%)	⬇️

---

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 1943d8e...52166b1. Read the comment docs.