Improve image inset resampling quality

[has2k1]

Summary

Image insets (inset_element with a PIL.Image or numpy.ndarray) previously
handed the full-resolution image to matplotlib's BboxImage and let it resample
to the inset's size at render time. matplotlib's resampler aliases badly on large
downscales — the regime an inset typically hits — producing soft, moiré-ridden
images.

This change pre-resizes the source to the inset's exact device-pixel footprint
with PIL and renders it pixel-for-pixel (interpolation="none"), choosing the
filter by scaling direction.

Resampling policy

direction	filter	why
downscale	LANCZOS	antialiases the reduction; matplotlib's resampler aliases on large downscales
upscale / 1:1	NEAREST	stays crisp; smooth filters (bicubic, LANCZOS) look hazy and ring on hard edges

Implementation notes

• Inputs are normalised to a pristine PIL.Image once, on construction
  (_to_pil_image); a float ndarray follows matplotlib's [0, 1] convention.
• The resize happens in _arrange_in_box, not draw(): the device-pixel size is
  only known after the layout engine finalises the inset's bbox at render time.
  draw() just creates the BboxImage; _arrange_in_box sets its data.

Baselines

Updated the affected image-inset baselines (image_aspect_fit_top_right,
image_aspect_fit_bottom, image_standalone) to reflect the new resampling.

[codecov]

Codecov Report

❌ Patch coverage is 91.30435% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 87.07%. Comparing base (d7b9533) to head (fecdbfb).
⚠️ Report is 6 commits behind head on main.

Files with missing lines	Patch %	Lines
plotnine/composition/_inset_image.py	91.30%	1 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1077      +/-   ##
==========================================
- Coverage   87.07%   87.07%   -0.01%     
==========================================
  Files         208      208              
  Lines       14341    14355      +14     
  Branches     1789     1791       +2     
==========================================
+ Hits        12488    12500      +12     
- Misses       1286     1287       +1     
- Partials      567      568       +1     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

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