REF: route groupby plotting through dedicated _python_apply_plot

[jbrockmendel]

Behavior-neutral refactor split off from GH-64660 (the GH-41090 group-key name-pinning deprecation).

Groupby plotting — the .plot accessor, .plot.<method>(), and .hist() — previously went through _python_apply_general, relying on the group key being pinned to each group's name attribute as a side effect of the general apply path. This routes it through a dedicated _python_apply_plot helper that labels each Series group with its key explicitly.

This is a pure refactor (no behavior change). It expresses plotting's actual need — give each Series group its key so plotting methods can use it for legend labels — directly, instead of piggybacking on the name-pinning side effect. It also stands as groundwork for GH-41090: once plotting no longer depends on the pin, that deprecation can be enforced without touching plotting.

_python_apply_plot is equivalent to the general path for plotting:

• Same (key, group) iteration (get_iterator is zip(result_index, _get_splitter(data)), exactly what apply_groupwise zips).
• not_indexed_same is hardcoded True — plotting always returns matplotlib objects, which are never index-like, so the general path already computed True here.
• Name labeling is narrowed to Series groups; DataFrame-group plotting takes legend labels from column names, never .name.

Test plan

• pandas/tests/plotting/test_groupby.py passes, including a new test_groupby_plot_series_with_legend locking in key-as-legend-label through the new path
• Full plotting + groupby suites pass (26534 passed, 0 failures)
• pre-commit and mypy clean (no new errors)

🤖 Generated with Claude Code