Add paired_raincloud() for repeated-measures connecting lines

[pengkodammaya]

Closes #56.

What

Implements the repeated-measures roadmap item as a dedicated function,
paired_raincloud() (per the design discussion on the issue — keeping
RainCloud's signature untouched). It draws a standard raincloud and connects
each subject's rain points across categories, making within-subject change
visible for paired / pre-post / longitudinal designs.

import ptitprince as pt

pt.paired_raincloud(
    x="condition", y="score", data=df, orient="h",
    id="subject",                                        # subject identifier column
    line_color="gray", line_alpha=0.3, line_width=0.5,   # line styling
)

Pre→post on 30 subjects: each thin line is one subject, so within-subject change is visible on top of the group distributions.

How

• paired_raincloud() wraps RainCloud() — it calls it for all the existing
  cloud/box/rain drawing (so styling stays identical and in one place), then adds
  the connecting lines on top. RainCloud itself is not modified at all.
• The rain is drawn by the existing stripplot; seaborn doesn't expose the
  jittered coordinates, so a private helper _draw_repeated_measures_lines()
  recovers them from the rain PathCollection offsets (the cloud is a
  PolyCollection and the box is patches/lines, so the rain is the only
  PathCollection added) and matches each point back to its subject using the
  within-category row order of data, which seaborn preserves.
• NaN measurements are filtered to keep that row-order matching aligned with the
  points seaborn actually drew, so a missing observation simply breaks that
  subject's line instead of misaligning everything.

Scope / limitations (v1)

• Supports the no-hue / no-dodge case (the overwhelmingly common one).
• hue / dodge are not accepted — with dodged sub-groups the
  point→subject mapping is ambiguous, so they raise a clear ValueError rather
  than silently drawing something wrong. Extending to dodged sub-groups is a
  sensible follow-up.

Backwards compatibility

Purely additive — a new function plus a private helper. RainCloud is untouched,
so existing behaviour is byte-for-byte unchanged.

RainCloud output is exactly as before.

Tests

Adds TestPairedRaincloud (+ a paired_data fixture in conftest.py):

• one line per subject, horizontal and vertical;
• a plain RainCloud draws no per-subject lines;
• NaN measurements don't error;
• hue / dodge raise ValueError;
• non-column-name inputs raise;
• line-style kwargs accepted; styling kwargs (move, point_size) forward to RainCloud.

All green locally on seaborn 0.13.2 / matplotlib 3.x.

Docs

• New function carries a full docstring (id / line_color / line_alpha / line_width).
• Adds a standalone repeated-measures tutorial notebook
  (tutorial_python/tutorial_repeated_measures.ipynb) using the existing Bristol
  AnxietyPaper dataset — which is already repeated-measures, so the feature
  drops straight in.

The feature on the tutorial's own AnxietyPaper data: each participant connected across the four emotion conditions.