Forward port fix scale domain no data #1619 (#1620)

* fix: set domain to (0, 1) when no data is present

Fixes #1614

If no data is present, a scale has a domain of (0, 1) but if we
remove the data (set it to an empty array) afterwards, we set it
to (-inf, +inf) which causes rendering issues and make the bqplot
figure unable to restore itself.
It is more consistent to behave like no data was passed in, and use
(0, 1) for the domain.

* Update snapshots

* Require bqscales recent

* Update Playwright Snapshots

---------

Co-authored-by: Maarten A. Breddels <maartenbreddels@gmail.com>
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>

setup.cfg

@@ -17,7 +17,7 @@ install_requires =
     traittypes>=0.0.6
     numpy>=1.10.4
     pandas>=1.0.0,<3.0.0
-    bqscales>=0.3.1,<0.4
+    bqscales>=0.3.3,<0.4
 classifiers =
     Framework :: Jupyter
     Framework :: Jupyter :: JupyterLab

test-environment.yml

@@ -9,7 +9,7 @@ dependencies:
     - traittypes >=0.0.6
     - numpy >=1.10.4,<2.0.0
     - pandas >=1.0.0,<2.0.0
-    - bqscales  >=0.3.1
+    - bqscales >=0.3.3
     - scipy
     - notebook
     - yarn =3

ui-tests/tests/notebooks/lines_update.ipynb

@@ -138,6 +138,26 @@
    "source": [
     "lines.labels_visibility = 'label'"
    ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "7444d361-f8f3-4641-88ef-fbe84450bfb3",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "lines.x, lines.y = [], []"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "481f09e7-f7a9-4393-8388-a608b216ef41",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "lines.x, lines.y = np.arange(len(y1)), [y1, y2, y3, y4]"
+   ]
   }
  ],
  "metadata": {