Use and depend on intersphinx_registry for doc building.

We try to still work if intersphinx_registry is not importable as it
might not be packaged yet, but it is now avail on conda forge, so it
should not be too hard.

docs/requirements.txt

@@ -9,4 +9,4 b' matplotlib'
 docrepr
 prompt_toolkit
 ipykernel
-stack_data
+intersphinx_registry

docs/source/conf.py

@@ -92,12 +92,36 b' exec('
 
 locals().update(config["sphinx"])
 
-intersphinx_mapping = config["intersphinx_mapping"]
-for k, v in intersphinx_mapping.items():
-    intersphinx_mapping[k] = tuple(
-        [intersphinx_mapping[k]["url"], intersphinx_mapping[k]["fallback"]]
+try:
+    from intersphinx_registry import get_intersphinx_mapping
+
+    intersphinx_mapping = get_intersphinx_mapping(
+        packages={
+            "python",
+            "rpy2",
+            "jupyterclient",
+            "jupyter",
+            "jedi",
+            "traitlets",
+            "ipykernel",
+            "prompt_toolkit",
+            "ipywidgets",
+            "ipyparallel",
+            "pip",
+        }
     )
 
+except ModuleNotFoundError:
+    # In case intersphinx_registry is not yet packages on current plaform
+    # as it is quite recent.
+    print("/!\\ intersphinx_registry not installed, relying on local mapping.")
+    intersphinx_mapping = config["intersphinx_mapping"]
+    for k, v in intersphinx_mapping.items():
+        intersphinx_mapping[k] = tuple(
+            [intersphinx_mapping[k]["url"], intersphinx_mapping[k]["fallback"]]
+        )
+
+
 # numpydoc config
 numpydoc_show_class_members = config["numpydoc"][
     "numpydoc_show_class_members"

pyproject.toml

@@ -63,6 +63,7 b' black = ['
 doc = [
     "docrepr",
     "exceptiongroup",
+    "intersphinx_registry",
     "ipykernel",
     "ipython[test]",
     "matplotlib",
@@ -70,7 +71,6 b' doc = ['
     "sphinx-rtd-theme",
     "sphinx>=1.3",
     "sphinxcontrib-jquery",
-    "stack_data",
     "tomli ; python_version<'3.11'",
     "typing_extensions",
 ]