Reject NaN pixels in render_images

[timtreis]

Summary

• render_images previously had asymmetric, silent NaN handling: single-channel substituted na_color via cmap.set_bad, multi-channel additively composited NaN as opaque black through per-channel cmaps that did not carry set_bad. Both paths hid upstream data issues (stitched gaps, padded crops, masked tissue, segmentation outputs).
• Both paths now raise ValueError early in _render_images with the element name, offending channel(s), and a fillna(...) hint.
• Integer-dtype channels are skipped (cannot contain NaN) by gating the check on np.issubdtype(layer.dtype, np.floating) before materializing, so uint8/uint16 microscopy images pay no cost.

Closes #628

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 77.66%. Comparing base (59da170) to head (08d9bbd).
⚠️ Report is 2 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #673      +/-   ##
==========================================
+ Coverage   77.63%   77.66%   +0.03%     
==========================================
  Files          11       11              
  Lines        3612     3626      +14     
  Branches      850      856       +6     
==========================================
+ Hits         2804     2816      +12     
  Misses        485      485              
- Partials      323      325       +2     

Files with missing lines	Coverage Δ
src/spatialdata_plot/pl/render.py	87.10% <100.00%> (+0.12%)	⬆️

... and 2 files with indirect coverage changes

🚀 New features to boost your workflow:

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