Refactor stage 2 aggregation for antialiased lines

[ianthomas23]

This PR refactors the second stage aggregation for antialiased lines so that the mathematical aggregation operations are passed into the line rendering functions from the Reduction classes, and in such a way that multiple reductions can be supported. There is no functional change from the end users point of view.

This is the final set of changes required before issue #1133 can be implemented for antialiased lines on the CPU.

Implementation details

• An antialias_stage_2 tuple is created from the supplied Reductions and passed into the line rendering functions. This tuple contains one item per Reduction, and each item is a 2-tuple of AntialiasCombination (an enum specifying the mathematical combination operator) and a zero_value which is the initial value of the aggregation (currently only np.nan but will be 0 for e.g. _sum_zero).

[codecov]

Codecov Report

Merging #1145 (3b43022) into master (e888ba3) will increase coverage by 0.03%.
The diff coverage is 94.26%.

@@            Coverage Diff             @@
##           master    #1145      +/-   ##
==========================================
+ Coverage   85.16%   85.20%   +0.03%     
==========================================
  Files          33       34       +1     
  Lines        7632     7732     +100     
==========================================
+ Hits         6500     6588      +88     
- Misses       1132     1144      +12     

Impacted Files	Coverage Δ
datashader/compiler.py	95.04% <88.88%> (-1.12%)	⬇️
datashader/glyphs/line.py	92.71% <93.03%> (-0.51%)	⬇️
datashader/reductions.py	85.84% <96.55%> (+0.48%)	⬆️
datashader/core.py	87.76% <100.00%> (-0.23%)	⬇️
datashader/data_libraries/dask.py	94.49% <100.00%> (ø)
datashader/data_libraries/dask_xarray.py	98.91% <100.00%> (ø)
datashader/data_libraries/pandas.py	100.00% <100.00%> (ø)
datashader/enums.py	100.00% <100.00%> (ø)
datashader/glyphs/area.py	79.82% <100.00%> (ø)
datashader/glyphs/points.py	88.29% <100.00%> (ø)
... and 4 more

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[jbednar]

It seems like this is necessary, and it doesn't make things appreciably worse, so go for it!