Generalize internal scaling operation

[mwaskom]

This PR updates most sns functions to use the transforms set on a pre-existing matplotlib axes such that statistics are computed in the visualized space. Previously, support for transformed axes was a little spotty, at most supporting log scales (but not similar scales, like symlog).

f, axs = plt.subplots(1, 4, figsize=(6, 3), sharey=True, layout="constrained")
plots = [sns.lineplot, sns.pointplot, sns.boxplot, sns.violinplot]
for ax, func in zip(axs.flat, funcs):
    ax.set_yscale("symlog")
    plot(diamonds, x="color", y="price", ax=ax)

This also fixes #3352, a bug introduced in the v0.13 dev series when jumping the gun on this functionality.

Arbitrary scales are also now handled properly on the orient axis of categorical plots:

f, ax = plt.subplots()
ax.set_xscale("symlog")
sns.boxplot(tips, x=10 ** tips["size"], y="total_bill", hue="sex", native_scale=True)

The code is fairly messy but, alas, that's rather unavoidable given the architecture underlying the function interface and was a major motivator for the development of the objects interface (and corresponding rearchitecting).

[codecov]

Codecov Report

Merging #3440 (44d4418) into master (2386036) will increase coverage by 0.00%.
The diff coverage is 100.00%.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #3440   +/-   ##
=======================================
  Coverage   98.32%   98.33%           
=======================================
  Files          77       77           
  Lines       24305    24335   +30     
=======================================
+ Hits        23899    23929   +30     
  Misses        406      406           

Files Changed	Coverage Δ
seaborn/_base.py	97.66% <100.00%> (+0.11%)	⬆️
seaborn/categorical.py	98.90% <100.00%> (-0.01%)	⬇️
seaborn/distributions.py	96.23% <100.00%> (-0.12%)	⬇️
seaborn/relational.py	99.69% <100.00%> (+<0.01%)	⬆️
tests/test_base.py	99.69% <100.00%> (+<0.01%)	⬆️
tests/test_categorical.py	99.13% <100.00%> (+<0.01%)	⬆️