Improve legend for categorical scatterplots

[mwaskom]

This ties up some loose ends from #2413 and #2447, which allowed them to apply a continuous mapping to the hue variable without upgrading the legend machinery to properly show a colorbar-like legend. Now that works as expected:

sns.catplot(diamonds, x="cut", y="price", hue="carat", jitter=.4, s=5)

The solution is a bit of a shortcut, simply making _CategoricalPlotterNew subclass _RelationalPlotter, where the relevant method for creating legend artists lives. Probably that method should be moved down to the base class level to keep a cleaner division between the modules (but also, the distributions module has a different approach altogether). Legends remain a mess.

[codecov]

Codecov Report

Merging #2828 (d2913f2) into master (fefd940) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master    #2828   +/-   ##
=======================================
  Coverage   98.04%   98.05%           
=======================================
  Files          68       68           
  Lines       21991    22014   +23     
=======================================
+ Hits        21561    21585   +24     
+ Misses        430      429    -1     

Impacted Files	Coverage Δ
seaborn/relational.py	99.69% <ø> (ø)
seaborn/categorical.py	98.32% <100.00%> (+0.08%)	⬆️
seaborn/tests/test_categorical.py	99.77% <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 fefd940...d2913f2. Read the comment docs.