markdown: Catch urllib3 exceptions which are raised during streaming.

requests transforms the base urllib3 exceptions into
requests.RequestExceptions -- but only within code that it is
running.  When parsing the streaming body in fetch_open_graph_image,
the read itself (inside lxml) may trigger urllib3 to raise its own
timeout error -- which escapes the current catch of
requests.RequestExceptions.

Catch both requests and urllib3 exceptions.

zerver/lib/markdown/__init__.py

@@ -44,6 +44,7 @@ import re2
 import regex
 import requests
 import uri_template
+import urllib3.exceptions
 from django.conf import settings
 from markdown.blockparser import BlockParser
 from markdown.extensions import codehilite, nl2br, sane_lists, tables
@@ -474,7 +475,7 @@ def fetch_open_graph_image(url: str) -> Optional[Dict[str, Any]]:
                     elif element.get("property") == "og:description":
                         og["desc"] = element.get("content")
 
-    except requests.RequestException:
+    except (requests.RequestException, urllib3.exceptions.HTTPError):
         return None
 
     return None if og["image"] is None else og