Test updates following release of datashader 0.14.1

[ianthomas23]

This PR fixes the HoloViews test failures which were identified in PR #5343. There were two sets of failures, both relating to HoloViews tests needing to be updated following changes released in Datashader 0.14.1. They were:

1. Improved trimesh rendering.
2. The default setting of rescale_discrete_levels=True in shade.

The former required changing the comparison arrays; I have visually checked all of these. The latter required setting rescale_discrete_levels=False and then the tests pass.

[codecov-commenter]

Codecov Report

Merging #5344 (08dddb3) into master (22db2d1) will decrease coverage by 0.00%.
The diff coverage is 78.57%.

@@            Coverage Diff             @@
##           master    #5344      +/-   ##
==========================================
- Coverage   88.12%   88.12%   -0.01%     
==========================================
  Files         300      300              
  Lines       61856    61862       +6     
==========================================
+ Hits        54513    54516       +3     
- Misses       7343     7346       +3     

Impacted Files	Coverage Δ
holoviews/tests/element/test_comparisonsimple.py	90.78% <50.00%> (-3.74%)	⬇️
holoviews/core/util.py	86.03% <100.00%> (+0.02%)	⬆️
holoviews/element/comparison.py	89.85% <100.00%> (ø)
holoviews/tests/operation/test_datashader.py	97.36% <100.00%> (+<0.01%)	⬆️

---

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 22db2d1...08dddb3. Read the comment docs.

[ianthomas23]

There are some extra test failures in CI where we are using equality for floating-point comparisons rather than allowing some tolerance. The actual problem is that we compare the types of two objects before deciding which comparison operator to choose and the code concludes that float and np.float64 are not the same types and drops to a default exact comparison. The solution is therefore to allow any two floating-point types to use the floating-point comparison rather than direct equality.

[philippjfr]

Thanks @ianthomas23!