fix(tests): align CI tests with recent security hardening (#31470)

Four recent security PRs landed on main with stale/missing test updates,
breaking 4 test shards on every subsequent PR's CI run:

- test_discord_bot_auth_bypass.py (PR #30742 c3caca658):
  DISCORD_ALLOWED_ROLES no longer bypasses _is_user_authorized.
  Inverted 3 tests to assert the new (correct) behavior: role config
  alone does NOT authorize at the gateway layer.

- test_msgraph_webhook.py (PR #30169 4ca77f105):
  adapter.is_connected is a @property, not a method. Test was calling
  it with () after the connect() change; TypeError: 'bool' is not
  callable. Removed the parens.

- test_feishu_approval_buttons.py (PR #30744 bdb97b857):
  Card-action callbacks now go through _allow_group_message
  authorization. 3 tests in TestCardActionCallbackResponse didn't
  populate adapter._allowed_group_users so the operator's open_id got
  rejected. Added the allowlist setup to each test, matching the
  existing pattern in test_returns_card_for_approve_action.

Also raise tolerance on test_wait_for_process_kills_subprocess_on_keyboardinterrupt:
the SIGTERM → 3s TimeoutStopSec → SIGKILL → reap chain can exceed 10s
under loaded xdist (40 workers). Bumped _wait_for_pgid_exit timeout
10→30s and worker join timeout 5→15s. Passes 100% in isolation
already; this just makes it tolerant of CI-host load.

Validation: 270/270 tests pass across the 5 affected files.

tests/tools/test_local_interrupt_cleanup.py

@@ -48,8 +48,14 @@ def _process_group_snapshot(pgid: int) -> str:
     ).stdout.strip()
 
 
-def _wait_for_pgid_exit(pgid: int, timeout: float = 10.0) -> bool:
-    """Wait for a process group to disappear under loaded xdist hosts."""
+def _wait_for_pgid_exit(pgid: int, timeout: float = 30.0) -> bool:
+    """Wait for a process group to disappear under loaded xdist hosts.
+
+    The cleanup chain is: SIGTERM → 3s TimeoutStopSec → SIGKILL → reap.
+    Under heavy xdist load (40 parallel workers, 6-shard CI), the full
+    sequence can exceed 10s. Default timeout is generous to avoid CI
+    flakes; in practice the wait returns in <1s on quiet hosts.
+    """
     deadline = time.monotonic() + timeout
     while time.monotonic() < deadline:
         if not _pgid_still_alive(pgid):
@@ -166,9 +172,11 @@ def test_wait_for_process_kills_subprocess_on_keyboardinterrupt():
         assert ret == 1, f"SetAsyncExc returned {ret}, expected 1"
 
         # Give the worker a moment to: hit the exception at the next poll,
-        # run the except-block cleanup (_kill_process), and exit.
-        t.join(timeout=5.0)
-        assert not t.is_alive(), "worker didn't exit within 5 s of the interrupt"
+        # run the except-block cleanup (_kill_process), and exit.  Under
+        # xdist load the SIGTERM → 3s wait → SIGKILL chain can take longer
+        # than 5s before the worker's join() returns; bumped to 15s.
+        t.join(timeout=15.0)
+        assert not t.is_alive(), "worker didn't exit within 15 s of the interrupt"
 
         # The critical assertion: the subprocess GROUP must be dead.  Not
         # just the bash wrapper — the 'sleep 30' child too. Under xdist load,