fix reproject in virtual product

[emmaai]

Reason for this pull request

• issue Need the spatial coords to carry correct spatial_ref and crs #1072
• odc.algo.xr_reproject crisp clean

Proposed changes

• use the implementation odc.algo.xr_reproject

• Closes Need the spatial coords to carry correct spatial_ref and crs #1072

• Tests added / passed

• Fully documented, including docs/about/whats_new.rst for all changes, note: nothing new, bug fix and better

[codecov]

Codecov Report

Merging #1073 (6b72a0b) into develop (6ee0462) will increase coverage by 0.21%.
The diff coverage is 25.00%.

@@             Coverage Diff             @@
##           develop    #1073      +/-   ##
===========================================
+ Coverage    93.14%   93.36%   +0.21%     
===========================================
  Files          102      102              
  Lines         9949     9927      -22     
===========================================
+ Hits          9267     9268       +1     
+ Misses         682      659      -23     

Impacted Files	Coverage Δ
datacube/virtual/impl.py	77.04% <25.00%> (+3.99%)	⬆️

---

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 6ee0462...6b72a0b. Read the comment docs.

[uchchwhash]

Issue #1072 is a serious oversight that I am responsible for, but this solution is unacceptable to me as this package IMHO should not have odc-tools as a dependency. Let me see if I can fix it.

[uchchwhash]

Please do not import odc.algo.

[Kirill888]

Agree with @uchchwhash odc.algo itself depends on datacube, we don't want to create dependency loops like that. Probably best to move some of the odc.algo code into datacube core and add proper tests for it. Or just fix current code. If one assumes that one time-slice fits into RAM without problems then you CAN have much simpler reprojection logic for Dask arrays and maybe for this application it is best to stick with that approach. Implementation still needs to be correct though.

[emmaai]

I'll close the pull request but stay on the branch for a while until #1072 fixed as everybody agrees on

[snowman2]

odc.algo itself depends on datacube, we don't want to create dependency loops like that. Probably best to move some of the odc.algo code into datacube core and add proper tests for it

Here is another alternative: opendatacube/odc-tools#58 😄