Fix FacetGrid.refline picking up hue_kws aesthetics like marker - #3979

Fixes

Fixes #3973

Problem

lmplot registers a marker aesthetic in FacetGrid.hue_kws (even when there is no hue mapping), and FacetGrid.map unconditionally injects every hue_kws entry into the mapped function. Since FacetGrid.refline draws its reference lines via self.map(plt.axvline / plt.axhline, ...), the marker rides along into a function that never asked for it, and axvline/axhline draw a dot at both ends of the reference line:

import seaborn as sns
import pandas as pd

df = pd.DataFrame({"x": [1.0, 2.0, 3.0], "y": [0.0, 1.0, 2.0]})
g = sns.lmplot(df, x="x", y="y")
g.refline(x=0, y=0)   # gray dots appear on both ends of the line

Change

Draw the reference lines directly on each facet axis instead of routing them through map(), so hue_kws aesthetics (markers, etc.) are never injected into axvline/axhline. The color=None hue-mapping behavior is preserved through _facet_color, which now resolves the per-hue palette color.

Validation

• New regression test test_refline_no_hue_kws asserts:
  • reference lines from lmplot().refline() carry no marker;
  • with a hue mapping, refline(color=None) still draws each line in its hue's palette color.
• tests/test_axisgrid.py: 124 passed.
• tests/test_regression.py: 52 passed, 9 skipped (unrelated).

Note: this change was prepared with AI assistance; the root-cause analysis and the approach (bypass map() in refline, keep map()'s hue_kws injection intact for data-mapped functions) were reviewed and validated locally against the seaborn test suite.