perf(color): map categorical shapes colour by code, not per-row hex parse

[timtreis]

Problem

ColorSpec.to_rgba — the matplotlib shapes fill + outline colour mapping — parses every row via to_rgba_array(list(color_vector)). For a categorical color_vector (K distinct hex strings over N rows) that is O(N) string parsing for O(K) actual colours.

Fix

• Categorical color_vector: parse the K categories once, gather by .codes (codes ≥ 0 — resolution fills NaN with an explicit na_color category, _color.py:682–685).
• Object color_vector (align_to_length pad / uniform na): pd.factorize(sort=False) on the distinct colours, then gather. (np.unique would sort the hex strings and is slower than the baseline at scale — deliberately avoided.)

Byte-identical

Verified main-vs-branch on real renders — categorical fill, categorical+outline, categorical+groups, none — max|diff| = 0. Unit test TestColorSpecToRgba locks to_rgba against the per-row to_rgba_array(list(...)) across all 5 color_vector variants (clean / na-category / filtered / none-uniform / object-after-align).

Scope & impact

• Matplotlib shapes path only. Continuous (source_vector is None) untouched; labels (map_array codes) and points (perf(render_points): keep categorical colour as codes+palette (memory at scale) #731) already use the compact form; datashader shapes don't call to_rgba.
• Function micro-benchmark: ~56× categorical, ~16× object, ~2.5× less peak on the colour array. End-to-end it is a small fraction of a shapes render (geometry build dominates post-perf(render_shapes): PathCollection of pre-built Paths (avoid per-shape Patch construction) #735) — this lands as a clean, low-risk, byte-identical reduction, not a headline end-to-end speedup.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 79.28%. Comparing base (55f4970) to head (f486614).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #737      +/-   ##
==========================================
+ Coverage   79.26%   79.28%   +0.02%     
==========================================
  Files          17       17              
  Lines        4596     4601       +5     
  Branches     1028     1029       +1     
==========================================
+ Hits         3643     3648       +5     
  Misses        603      603              
  Partials      350      350              

Files with missing lines	Coverage Δ
src/spatialdata_plot/pl/_color.py	69.30% <100.00%> (+0.24%)	⬆️

🚀 New features to boost your workflow:

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