Preserve Categorical dtype for color_vector with non-unique colors

[timtreis]

Summary

• pd.Categorical.map() silently demotes to object dtype when mapped values aren't unique (e.g. two categories share the same color via .uns colors, or >103 categories all mapping to grey). This one-line fix wraps the result back in pd.Categorical.
• Labels: _map_color_seg now always takes the fast code-based label2rgb path (2-3x faster per render pass, compounding with fill+outline)
• Shapes/Points: datashader color_key is now correctly built from assigned colors instead of being silently None
• Also adds explicit na_action="ignore" to silence the FutureWarning from pandas >=2.1

Closes #469
Closes #540

Details

The root cause is in _set_color_source_vec where color_source_vector.map(color_mapping) can return an Index with object dtype instead of a Categorical when the color mapping contains duplicate values. This happens because pandas can't represent non-unique values as categorical categories.

Downstream effects of the object dtype:

• _map_color_seg falls into Case D (slow path): label2rgb processes N colors per instance instead of K unique category colors
• Shapes/points datashader: color_key check type(color_vector) is Categorical fails → assigned colors are ignored
• Labels rasterize path: hits assert color_source_vector is None (latent crash when rasterization drops labels)

The fix is verified to produce pixel-identical output at all tested scales (20–5000 cells).

Test plan

• Pixel-identical output verified for Categorical vs object-dtype paths (small/medium/large)
• NaN values correctly preserved (code=-1 → background)
• End-to-end render_labels with non-unique .uns colors completes without error
• Existing test suite passes

🤖 Generated with Claude Code

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 83.96%. Comparing base (c538e7f) to head (75d3dfa).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #542   +/-   ##
=======================================
  Coverage   83.96%   83.96%           
=======================================
  Files           9        9           
  Lines        2595     2595           
=======================================
  Hits         2179     2179           
  Misses        416      416           

Files with missing lines	Coverage Δ
src/spatialdata_plot/pl/render.py	88.80% <ø> (ø)
src/spatialdata_plot/pl/utils.py	79.30% <100.00%> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.