fix: render unannotated shapes with na_color instead of dropping them

[timtreis]

Closes #710.

What

render_shapes colored by a table column silently dropped shapes whose instance had no row in the table (_join_table_for_element(how="inner")). This renders them with na_color instead — consistent with the points (how="left" merge) and labels (raster + na_color) paths, and with the scanpy/squidpy convention.

# 20 shapes, table annotates 12
sdata.pl.render_shapes("shapes", color="val").pl.show()
# before: 12 shapes drawn (8 silently disappear)
# after:  20 shapes drawn (8 unannotated -> na_color)

How

One-line behavior change: _join_table_for_element uses how="left" instead of how="inner", so every shape survives the join. The per-shape color lookup (_extract_color_column, from #709, reindexed to the element) then yields NaN → na_color for the unannotated shapes via the existing NaN-handling path (which also emits the existing "Found N NaN values… colored with na_color" warning). No new color logic.

Scope / safety

• Behavior-only, deliberately separate from the performance work — minimal diff (utils.py + a test). The structural join still runs; removing it (the remaining color-copy speedup) is a separate follow-up.
• Fully-annotated data is unaffected — verified pixel-identical (inner == left when nothing is dropped) for continuous, categorical, and gene coloring. So existing visual baselines should not change; only partial-annotation fixtures (if any) change — those would be a deliberate, reviewable baseline update.
• Builds on perf(color): extract one aligned column instead of copying the whole table #709 (_extract_color_column), which is already on main.

Test

Adds test_render_shapes_unannotated_shapes_get_na_color_not_dropped — a real regression test: it fails on the old code (expected all 20 shapes rendered, got 12 (unannotated dropped)) and passes on the fix (20 rendered, 8 in na_color). There was previously no test covering this behavior, which is why the inconsistency went unnoticed.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 76.39%. Comparing base (8d74219) to head (6fb510f).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #711      +/-   ##
==========================================
- Coverage   76.42%   76.39%   -0.03%     
==========================================
  Files          14       14              
  Lines        4330     4330              
  Branches     1006     1006              
==========================================
- Hits         3309     3308       -1     
  Misses        664      664              
- Partials      357      358       +1     

Files with missing lines	Coverage Δ
src/spatialdata_plot/pl/render.py	87.08% <ø> (ø)
src/spatialdata_plot/pl/utils.py	69.07% <ø> (-0.05%)	⬇️

🚀 New features to boost your workflow:

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