introduce set_legend_title() and make legend object available via get_legend_obj()

[berndreinhold]

Introduced set_legend_title() which modifies the legend title without having to recreate the legend as in move_legend() - it is thereby a bit more lightweight. It also takes a prop-argument to modify e.g. title fonts.

For more elaborate modifications get_legend_obj() can be used to get a legend object. get_legend_obj() contains an extraction of code found within move_legend(), so that it can be called both from move_legend() and set_legend_title() or by itself.

This is related to:

• Plot legend needs more customizability #2994 and
• How to do various things in the next gen Seaborn #2978 (comment)
  and potentially addresses them, by making the legend object available.

Introduced tests for both functions, modeled after test_move_legend*()
"make tests" gave very similar results compared to results before the modification:
1943 passed, 14 skipped, 7 xfailed, 652 warnings

[mwaskom]

Hi, I don't see how this addresses the two linked issues, which have to do with the Plot interface. Modification of legend titles on existing matplotlib objects is already relatively straightforward, e.g.

ax = sns.histplot(penguins, x="bill_length_mm", hue="species")
ax.legend_.set_title("Kind of penguin")

[berndreinhold]

Thanks for pointing out the accessibility of legend for axes (and FacetGrid). And I had missed that the reference to Plot, was specifically to https://seaborn.pydata.org/generated/seaborn.objects.Plot.html. This renders this PR obsolete.

[mwaskom]

Also while the Plot legend customizability remains lacking in general, it's actually already possible to control the legend title through Plot.label:

(
    so.Plot(penguins, x="bill_length_mm", color="species")
    .add(so.Bars(), so.Hist())
    .label(color="Kind of penguin")
)

[berndreinhold]

Thanks for your quick answers!