fix(show): resolve axes/panels mismatch for multi-CS elements (#749)

[timtreis]

Summary

Fixes #749. pl.show(ax=single_ax) raised

ValueError: Mismatch between number of matplotlib axes objects (1) and number of panels (2).

when an element carried transformations to several coordinate systems that render the same elements (e.g. visium's "<cs>" and "<cs>_downscaled_lowres"). filter_by_coordinate_system can't strip the extra transformation (upstream scverse/spatialdata#176), so auto-detection produced one panel per coordinate system and a single user axes no longer matched.

PR #580's strict_cs narrowing doesn't help here: both coordinate systems contain the same element, so neither is filtered. The two panels are redundant — identical content, differing only by a scale transform.

Fix

In _resolve_coordinate_systems, after the existing strict_cs step, deduplicate auto-detected coordinate systems by their renderable-element set when axes are supplied: keep the first of each distinct set so a single axes is satisfied, and emit a UserWarning naming the dropped duplicates and pointing to coordinate_systems=. Genuinely distinct coordinate systems keep more sets than axes and still fall through to the existing mismatch ValueError.

Scoped to the ax-provided, auto-detected path only:

• This path previously raised, so making it render breaks nothing that worked before (backward compatible, as the reporter requested).
• The no-ax multi-panel behaviour is untouched.
• Explicit coordinate_systems= is untouched (cs_was_auto is False).

Tests

• test_single_ax_auto_cs_redundant_duplicates_resolved — redundant duplicate CS + single ax now renders one panel and warns (replaces Fix axes mismatch when elements have multi-CS transformations #580's ..._unresolvable_raises, whose "raise with hint" stopgap is exactly what pl.show() leading to ValueError due to mismatch between number of axes and number of panels #749 supersedes).
• test_single_ax_auto_cs_distinct_elements_raises — an image-only CS and a shapes-only CS on one axes are genuinely distinct panels and still raise.

Non-visual suite passes locally; the only failures are the known CI-only test_plot_* visual-baseline comparisons. No baseline changes expected (new path exercised only by the non-visual regression tests).

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 80.17%. Comparing base (87beb7f) to head (ab0f3cb).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #784      +/-   ##
==========================================
+ Coverage   80.11%   80.17%   +0.05%     
==========================================
  Files          18       18              
  Lines        4869     4882      +13     
  Branches     1081     1085       +4     
==========================================
+ Hits         3901     3914      +13     
  Misses        604      604              
  Partials      364      364              

Files with missing lines	Coverage Δ
src/spatialdata_plot/pl/basic.py	83.95% <100.00%> (+0.65%)	⬆️

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

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