fix: improve steinbock docstring accuracy and RST formatting

- Move notes into a proper NumPy `Notes` section (fixes RST render)
- Fix typos: andata→anndata, adviced→advised
- Correct adata.obs.images→adata.obs.image (verified from source)
- Replace plain URL with RST hyperlink syntax
- Clarify that only one masks dir is used at a time (controlled by labels_kind)
- Specify *.ome.tiff image format (from SteinbockKeys.IMAGE_SUFFIX)
- Explain the images.csv/--info mechanism with required columns and correct
  link to file types docs instead of preprocessing docs

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: LucaMarconato

.gitignore

@@ -34,6 +34,8 @@ __pycache__/
 _version.py
 node_modules/
 .code-workspace
+.asv/
+uv.lock
 
 # test datasets (e.g. Xenium ones)
 data/

src/spatialdata_io/readers/steinbock.py

@@ -33,24 +33,6 @@ def steinbock(
 
         - `Steinbock pipeline  <https://bodenmillergroup.github.io/steinbock/latest/>`_.
 
-    Current mandatory outputs of the Steinbock toolkit
-    ----------
-    - one comprehensive anndata object labeled as 'cells.h5ad' in the main working directory
-    - an 'ome' folder containing the steinbock ome.tiff exported tiff images
-    - masks directory labeled as either 'masks_deepcell' or 'masks_ilastik'
-
-    Users have to carefully construct the andata object with steinbock using the images.csv table.
-    Constructing the anndata object without the images.csv table will leave the anndata without
-    'adata.obs.images' which will subsequently result in an error (see line 99)
-    Therefore, currently users are adviced to start their steinbock analyses from .mcd files.
-    A possible workaround for this would be to handcraft the images.csv table according to steinbock format
-    if users want to start from tiff images in steinbock
-    (https://bodenmillergroup.github.io/steinbock/latest/cli/preprocessing/)
-
-    The andata tables, masks and ome.tiffs have to be named accordingly so the
-    spatialdata.io steinbock reader correctly assigns all instances.
-
-
     Parameters
     ----------
     path
@@ -65,6 +47,29 @@ def steinbock(
     Returns
     -------
     :class:`spatialdata.SpatialData`
+
+    Notes
+    -----
+    Mandatory outputs of the Steinbock pipeline required by this reader:
+
+    - ``cells.h5ad``: the AnnData object in the main working directory
+    - ``ome/``: directory containing the steinbock OME-TIFF images (``*.ome.tiff``)
+    - ``masks_deepcell/`` or ``masks_ilastik/``: the masks directory, selected via the
+      ``labels_kind`` parameter (only one is used at a time)
+
+    When exporting to AnnData with ``steinbock export anndata``, the ``--info`` option
+    (default: ``images.csv``) controls whether image metadata is embedded. When present,
+    steinbock copies the ``image`` column from ``images.csv`` into ``adata.obs.image`` for
+    every cell; this reader requires that column to map cells to their source image. Without
+    it, ``adata.obs.image`` will not exist and the reader will fail. ``images.csv`` is
+    generated automatically by ``steinbock preprocess imc images`` when starting from
+    ``.mcd`` files. Users who start from TIFF images must hand-craft ``images.csv`` with at
+    least the columns ``image``, ``width_px``, ``height_px``, and ``num_channels`` as
+    described in the
+    `Steinbock file types documentation <https://bodenmillergroup.github.io/steinbock/latest/file-types/>`_.
+
+    All AnnData tables, masks, and OME-TIFFs must follow the steinbock naming conventions so
+    that the reader can correctly match all instances.
     """
     path = Path(path)