Add support for weights in objects.Est

[mwaskom]

This PR enhances seaborn.objects.Est to support a weighted mean, with confidence intervals:

diamonds = sns.load_dataset("diamonds")
(
    so.Plot(diamonds, x="color", y="price")
    .add(so.Dot(), so.Est(), weight="carat")
    .add(so.Range(), so.Est(), weight="carat")
)

(I don't know that this plot makes any particular sense, but it demonstrates how you'd use it...)

It is using np.average under the hood.

Currently support is limited to func="mean", errorbar=("ci", ...). There may be other estimators / error bar methods were it makes sense to compute a weighted version (e.g. weighted median/percentiles) although we would need to implement those in seaborn as I don't believe there's an implementation in our dependencies. I think this is fine and probably satisfies the majority of use cases; we can expand support later if there's enough interest.

(As a side note, we should probably error when func != "mean" and errorbar=("se", ...) since we don't generically compute the parameteric standard error of whatever estimator is passed).

See previous discussion in #3563, which is about the function interface. I actually think that it will be relatively straightforward to add first-class support in the function interface too on top of these changes. I might follow up here or maybe do that in another PR...

[codecov]

Codecov Report

Merging #3580 (5225101) into master (63d91bf) will increase coverage by 0.18%.
Report is 10 commits behind head on master.
The diff coverage is 100.00%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #3580      +/-   ##
==========================================
+ Coverage   98.34%   98.52%   +0.18%     
==========================================
  Files          75       75              
  Lines       24625    24622       -3     
==========================================
+ Hits        24217    24259      +42     
+ Misses        408      363      -45     

Files	Coverage Δ
seaborn/_statistics.py	99.69% <100.00%> (+0.02%)	⬆️
seaborn/_stats/aggregation.py	97.61% <100.00%> (+0.11%)	⬆️
tests/_stats/test_aggregation.py	100.00% <100.00%> (ø)
tests/test_statistics.py	100.00% <100.00%> (ø)

... and 12 files with indirect coverage changes

[mwaskom]

Going to merge this and then follow up on the function interface in a separate PR.