Fix version discovery for clouds with int project_ids

On a cloud that has inaccessible version discovery documents AND uses
integer project ids, the discovery fallback logic can fail because the
project id parses as a (very large) version.

Check to see that the url segment in the fallback code begins with a v,
so that we're only attempting to parse versions from actual candidate
segments.

Closes-Bug: #1806109
Change-Id: Id90b3b9e4852494a4678b0a9bb67362babdc971c

keystoneauth1/discover.py

@@ -473,6 +473,12 @@ def _version_from_url(url):
     url = urllib.parse.urlparse(url)
     for part in reversed(url.path.split('/')):
         try:
+            # All integer project ids can parse as valid versions. In URLs
+            # all known instances of versions start with a v. So check to make
+            # sure the url part starts with 'v', then check that it parses
+            # as a valid version.
+            if part[0] != 'v':
+                continue
             return normalize_version_number(part)
         except Exception:
             pass

keystoneauth1/tests/unit/test_discovery.py

@@ -1257,3 +1257,8 @@ class EndpointDataTests(utils.TestCase):
         self.assertEqual(exp, str(epd))
         # Works with implicit stringification
         self.assertEqual(exp, "%s" % epd)
+
+    def test_project_id_int_fallback(self):
+        bad_url = "https://compute.example.com/v2/123456"
+        epd = discover.EndpointData(catalog_url=bad_url)
+        self.assertEqual((2, 0), epd.api_version)