perf(images): materialize rasterized image once to avoid per-channel re-warp

[timtreis]

What

render_images re-materializes the lazy rasterized image once per channel, on every pass, so the affine warp runs N_passes × N_channels times per render. This materializes it once.

Root cause

spatialdata.rasterize returns a lazy dask array (~5 ms). _render_images then reads it per channel in several passes — the NaN check, the per-channel compositing (img.sel(c=ch).values), and the draw — each of which recomputes the entire dask graph (the order=0 affine warp + concatenate3). So an 8-channel render pays for the warp roughly passes × 8 times.

Measured (synthetic 8ch float32, chunked like the real pipeline):

step	time
_rasterize_if_necessary (lazy)	5 ms
full pl.show (8ch 6000²)	10 245 ms
per-channel .sel(c=ch).values	2544 ms
single .compute() of the same array	320 ms (8.0× = channel count)

Fix

Materialize the rasterized array once before returning:

image = rasterize(...)
if hasattr(image.data, "compute"):
    image = image.copy(data=image.data.compute())

Results

End-to-end: 10 457 ms → 508 ms (20.6×) at 8ch 6000²; 15.4× at 4000². Single-channel is neutral. The datashader path already .compute()s, so it is unaffected.

Why this is safe

• Pixel-identical output (np.array_equal == True) — transform, coords, and channel labels preserved — so no visual baselines change (that's the correctness proof: existing image/label tests stay green unchanged).
• Memory-bounded: the output of _rasterize_if_necessary is always display-sized (rasterized to ~target, or returned only when ≤ target+100 per axis), so the eager compute is ~15 MB at 700²×8.
• Covers images and labels (both call _rasterize_if_necessary).

Closes #707.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 50.00000% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 76.36%. Comparing base (2c8803a) to head (46464e7).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
src/spatialdata_plot/pl/utils.py	50.00%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #708      +/-   ##
==========================================
- Coverage   76.40%   76.36%   -0.04%     
==========================================
  Files          14       14              
  Lines        4314     4316       +2     
  Branches     1003     1004       +1     
==========================================
  Hits         3296     3296              
  Misses        663      663              
- Partials      355      357       +2     

Files with missing lines	Coverage Δ
src/spatialdata_plot/pl/utils.py	68.96% <50.00%> (-0.07%)	⬇️

🚀 New features to boost your workflow:

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