fix: tile-invariant featurization (positions, truncation, multiscale, granularity) + lower output memory

[timtreis]

Tiled calculate_image_features (and calculate_tiling_qc) could silently return values that depend on tile_size:

• Positions were tile-local. Center_X/Y, BoundingBox*, Location_*, centroid-* are now in the labels' pixel grid.
• Asymmetric cells were truncated. Tiles now crop to the union of their owned cells' bounding boxes + 1 px, so no cell is cut. This also stops one huge object from inflating every tile's read.
• Multiscale labels dropped small cells. Centroids came from the coarsest scale only; the scan now uses the requested scale.
• Granularity drifted with tiling. Tiles that compute it are aligned to its 16 px sampling grid and padded by 256 px. On real Xenium nuclei, Spearman vs an untiled run for bins 1-5 goes from 0.77-0.90 to 0.94-0.997.

Output assembly no longer copies the full table three times: ~3x lower peak memory and ~27x faster on 1e8 values, with identical output.

# same values for any tile_size (except the coarsest granularity bins, which depend on context beyond any tile)
a = sq.experimental.im.calculate_image_features(sdata, image_key="img", labels_key="cells", tile_size=100_000, inplace=False)
b = sq.experimental.im.calculate_image_features(sdata, image_key="img", labels_key="cells", tile_size=1024, inplace=False)

[codecov]

Codecov Report

❌ Patch coverage is 95.77465% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 78.51%. Comparing base (8286274) to head (ba04760).

Files with missing lines	Patch %	Lines
...uidpy/experimental/im/_calculate_image_features.py	98.07%	1 Missing ⚠️
src/squidpy/experimental/im/_tiling.py	93.75%	0 Missing and 1 partial ⚠️
src/squidpy/experimental/tl/_tiling_qc.py	66.66%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1295      +/-   ##
==========================================
+ Coverage   78.44%   78.51%   +0.07%     
==========================================
  Files          63       63              
  Lines        9532     9537       +5     
  Branches     1594     1593       -1     
==========================================
+ Hits         7477     7488      +11     
+ Misses       1489     1486       -3     
+ Partials      566      563       -3     

Files with missing lines	Coverage Δ
...uidpy/experimental/im/_calculate_image_features.py	89.93% <98.07%> (+0.50%)	⬆️
src/squidpy/experimental/im/_tiling.py	90.64% <93.75%> (+1.57%)	⬆️
src/squidpy/experimental/tl/_tiling_qc.py	70.25% <66.66%> (-0.32%)	⬇️

... 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.