Be a lot more conservative in _github_submodule_required (#708)

This function is used to decide whether we can use github directly to
download the code without cloning the repository. That optimization only
works if the repository doesn't have submodules.

Before this commit, if github returned any status code that wasn't a 2xx
or 3xx, we would consider that the repository didn't have any submodule.
This is fairly obviously wrong as a transient server error on the github
side for example would mean that we'd assume that the repository doesn't
have any submodules.

We now check for HTTPErrors and only consider that a repository doesn't
have any submodule iif the status is 404 otherwise, something went
wrong, log it and fallback to the non optimized path.

I also kept the bare exception but added some log to it, just in case.

Author: Eijebong

src/taskgraph/run-task/fetch-content

@@ -679,8 +679,15 @@ def _github_submodule_required(repo: str, commit: str):
     try:
         status_code = urllib.request.urlopen(url).getcode()
         return status_code == 200
-    except:
-        return False
+    except urllib.error.HTTPError as e:
+        if e.status == 404:
+            return False
+        # If we get a non 2xx status code that isn't a 404, something has gone horribly wrong on the github side, log it and return True
+        log("Got {} from github while checking for submodules in {} which was unexpected. Cannot check whether the repo has submodules or not".format(e.status, repo))
+        return True
+    except Exception as e:
+        log("Got an unexpected `{}` exception while checking for submodules in {}. Cannot check whether the repo has submodules or not".format(e, repo))
+        return True
 
 
 def git_checkout_archive(