Add the concept of a "layer legend" to so.Plot

[mwaskom]

This introduces an alternate way to provide information about what different marks or mark properties mean. It is similar to how label= is used when passed to a specific matplotlib artist.

For example, you can use it to gloss marks that represent different kinds of values:

(
    so.Plot(tips, x="total_bill", y="tip")
    .add(so.Dots(), label="Data")
    .add(so.Line(), so.PolyFit(), label="Model")
)

It is also useful for plotting multiple variables in cases where it would be awkward to reshape the data source to long-form:

(
    so.Plot(tips, x="size")
    .add(so.Dots(color="C1"), so.Jitter(), y="total_bill", label="Bill")
    .add(so.Dots(color="C2"), so.Jitter(), y="tip", label="Tip")
    .label(legend="Source")
)

Closes #3046

[codecov]

Codecov Report

Merging #3456 (f9aa345) into master (0a64c1a) will increase coverage by 0.00%.
The diff coverage is 100.00%.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #3456   +/-   ##
=======================================
  Coverage   98.32%   98.33%           
=======================================
  Files          75       75           
  Lines       24408    24438   +30     
=======================================
+ Hits        24000    24030   +30     
  Misses        408      408           

Files Changed	Coverage Δ
seaborn/_marks/base.py	98.48% <ø> (ø)
seaborn/_core/plot.py	98.29% <100.00%> (+0.02%)	⬆️
tests/_core/test_plot.py	98.82% <100.00%> (+0.01%)	⬆️