Avoid unnecessary copies in snippets for file upload

Prior to this commit, we recommended extracting a string representation
of the uploaded file with:

uploaded_file.tobytes().decode("utf-8")

This causes a copy of the memoryview to be created when it's converted
to bytes. This copy is immediately discarded when it is decoded.
Switching to:

codecs.decode(uploaded_file, encoding="utf-8")

means that we apparently don't create the intermediate copy.

This commit also adds an example of using the uploaded file in an image
widget.

Author: pbugnion

docs/source/examples/Widget List.ipynb

@@ -1069,10 +1069,11 @@
     "#=> b'This is the content of example.txt.\\n'\n",
     "```\n",
     "\n",
-    "To get the contents as a string:\n",
+    "You can get the contents as a string if the file is a text file (more precisely, if the file can be decoded as UTF-8):\n",
     "\n",
     "```python\n",
-    "uploaded_file[\"content\"].tobytes().decode(\"utf-8\")\n",
+    "import codecs\n",
+    "codecs.decode(uploaded_file[\"content\"], encoding=\"utf-8\")\n",
     "#=> 'This is the content of example.txt.\\n'\n",
     "```\n",
     "\n",
@@ -1086,7 +1087,15 @@
     "To convert the uploaded file into a Pandas dataframe, you need to convert it to a StringIO object:\n",
     "\n",
     "```python\n",
-    "pd.read_csv(io.StringIO(uploaded_file[\"content\"].tobytes().decode(\"utf-8\")))\n",
+    "import io\n",
+    "import pandas as pd\n",
+    "pd.read_csv(io.StringIO(codecs.decode(uploaded_file[\"content\"], encoding=\"utf-8\")))\n",
+    "```\n",
+    "\n",
+    "If the uploaded file is an image, you can visualize it with an image widget:\n",
+    "\n",
+    "```python\n",
+    "widgets.Image(value=uploaded_file[\"content\"].tobytes())\n",
     "```\n",
     "\n",
     "<div class=\"alert alert-info\">\n",