[newmana]

• Memory usage is reduced through chunking in cci.run - CCI scoring and permutation
  testing ~5,000,000 bins @ 20GB in ~20 minutes.
• SpatialData support added with a new st.io package. Supporting:
  atera, xenium, visium_hd and visium. Start of
  spatial data native plotting with st.pl.sdata_plot and
  st.pp.highly_variable_genes.
• Spatial trajectory and other APIs use cluster labels strings
  (cluster="6").
• Upgraded ligand-receptor databases to connectomeDB2025 (human and mouse).
• Loading via load_lrs is deprecated for lr_pairs with
  connectomeDB2020_put removed. CCI and GO results change with the
  new databases.
• New datasets available to download via visium_hd_sge and atera_sge
  and return SpatialData objects.
• AnnData backwards compatibility available through st.io.attach_spatial.
• CCI results are now a LR table (AnnData) returned by st.tl.cci.run.
  A spots x LR pairs with summaries, scores and p-values in var.
  Downstream functions and plots now take lr_results.
• Support for microns through microns_per_pixel,
  st.utils.microns_to_pixels, pixels_to_microns and
  create_grid(bin_um=).
• SME computes weights only between nearby spots and data is kept sparse
  where possible.
• extract_feature supports ConvNeXt and takes weights=.

API and Bug Fixes:

• st.tl.cci.run(use_label=...) never applied cell heterogeneity weighting:
  it read the cell-type data from the wrong key, found nothing, and weighted
  every spot 1, so results were identical with and without use_label. It
  now weights by heterogeneity from cell-type proportions in uns[use_label]
  or per-cell labels in obs[use_label], and takes cell_prop_cutoff.
  CCI scores and significance change for any run that used use_label.
• run_cci changes: p-values calculated via permutation calculation
  via are (k+1)/(N+1) and can nolonger be 0, cutoff is now
  p < p_cutoff, and permutations are seeded through random_state.
• spot_mixtures matches cell types by exact name.
• n_spots in the LR summary counted spots with a score above min_expr,
  while the min_spots filter counted spots with a score above 0, so the
  reported count was not the one the filter used. Both now use score > 0;
  n_spots goes up for weakly scoring pairs.
• st.tl.cci.grid dropped cells that fell on the last row or column of the
  grid or exactly on a square boundary. Every cell now lands in exactly one
  square, so cell counts, summed expression and label proportions change,
  mostly in edge squares.
• lr_plot always drew lr_sig_scores whatever use_label asked for,
  so a plot labelled p_adjs or neg_log10_p_adjs showed the masked
  score instead. It now draws the statistic requested. lr_plot and
  ccinet_plot also draw on a fig or ax you pass in rather than
  opening a new figure, so they can go in subplots.
• lr_plot / ccinet_plot use the fig / ax passed in rather
  than creating their own.
• Images in uns["spatial"] are now uint8 (0-255), not floats in [0, 1].
• extract_feature tiles were fed as floats between 0-255 instead of 0-1.
  Tiles are now read as uint8, so X_tile_feature and X_morphology
  change.
• concatenate_spatial_adata: the x and y offsets added to obsm["spatial"]
  for each tile were swapped, so spots landed on the wrong tile unless the tile
  grid was symmetric. Tiles also keep their aspect ratio instead of being
  stretched to fixed_size, each sample is scaled by its own scale factor
  rather than the first sample's, and fixed_size is (height, width) like a
  NumPy shape.
• read_merfish works again with new notebook.
• read_10x reads 16-bit / signed PNGs and reports missing images clearly.
• add.labels / add_loupe_clusters match by barcode
• add.positions reads Space Ranger 2.0+ files
• add.parsing assigned spots the wrong coordinates (they were written in a
  different order from the spots in obs), so every plot and anything using
  obsm["spatial"] (neighbours, SME, CCI) was wrong. Each spot now gets its
  own position.
• SME fits each axis separately when finding neighbours.
• Spatial trajectory: pseudotime orients by median pseudotime (paths may
  differ), shortest_path_spatial_paga returns labels, plus fixes to
  local_level, weight_optimizing_local, de_transition_plot and the
  trajectory plots.
• Upgraded to current matplotlib (3.11.0).
• qc_plot no longer makes a dense copy.
• stlearn.wrapper removed.
• cci.grid vectorised and nolonger takes n_cpus.
• obsm["spatial"] is in full-resolution pixels (was image pixels), matching
  the rest of stLearn.
• SME outputs are sparse when X is sparse; no weight matrices in uns.
• Removed: extract_feature(cnn_base="xception"),
  deconvolution_plot(data_alpha=), local_plot(cluster=),
  Spatial(basis=).
• New defaults: trajectory node_size=60 (area);
  check_trajectory(size_spatial=7, img_key=None).