PERF: CRS - use WKT for cache and lazy load EPSG code

[snowman2]

Port of opendatacube/datacube-core#1322

[codecov]

Codecov Report

Base: 98.19% // Head: 98.14% // Decreases project coverage by -0.05% ⚠️

Coverage data is based on head (d5779b5) compared to base (9282c21).
Patch coverage: 89.47% of modified lines in pull request are covered.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop      #70      +/-   ##
===========================================
- Coverage    98.19%   98.14%   -0.06%     
===========================================
  Files           23       23              
  Lines         3768     3768              
===========================================
- Hits          3700     3698       -2     
- Misses          68       70       +2     

Impacted Files	Coverage Δ
odc/geo/crs.py	99.16% <89.47%> (-0.84%)	⬇️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[Kirill888]

@snowman2 thanks for this, but I'm not too sure this is the right way to approach this problem. I think rasterio.crs.CRS is a common enough input type and we should handle it in the _make_crs cached path, rather than handling it in a "generic" section. rasterio.crs.CRS is a hashable object so can be used as a key directly, only issue is that I want odc-geo to be independent from rasterio, so it might be a bit trickier to implement because of that.

I'm guessing that a performant way to handle rasterio.crs.CRS object would be to query .is_epsg_code property before attempting to call .to_epsg, also with those things going through cache performance impact will be reduced.