test: re-enable the OAuth router and rate limiter TTL suites

.github/workflows/test.yml

@@ -67,9 +67,7 @@ jobs:
             # Currently failing.
             --ignore=src/tests/api_gateway/test_app_startup_catalog.py
             --ignore=src/tests/api_gateway/test_config_modes.py
-            --ignore=src/tests/api_gateway/test_oauth_router.py
             --ignore=src/tests/api_gateway/test_oidc_role_mapping.py
-            --ignore=src/tests/api_gateway/test_rate_limiter_ttl.py
             --ignore=src/tests/cli/test_dashboard_config.py
             --ignore=src/tests/common/test_cli_providers.py
             --ignore=src/tests/common/test_cluster_naming.py

src/tests/api_gateway/test_oauth_router.py

@@ -16,7 +16,7 @@
 from httpx import ASGITransport, AsyncClient
 
 
-def _build_app(monkeypatch, *, auth_provider="external", base="https://auth.example.test",
+def _build_app(monkeypatch, *, auth_provider="oidc", base="https://auth.example.test",
                redirect_uri="https://app.example.test/auth/callback",
                client_id="inferiallm-dashboard") -> FastAPI:
     """Wire the oauth_router into a clean FastAPI app with patched settings."""

src/tests/api_gateway/test_rate_limiter_ttl.py

@@ -48,9 +48,15 @@ async def test_is_allowed_creates_bucket(self):
     @pytest.mark.asyncio
     async def test_rate_limit_exhaustion(self):
         limiter = InMemoryRateLimiter()
-        # Exhaust all tokens
+        # The defaults refill 166 tokens a second, which is faster than the
+        # loop below drains 1000 of them, so the bucket never empties. Both
+        # are set here to measure exhaustion rather than loop speed.
+        limiter.burst_size = 5
+        limiter.requests_per_minute = 0
+
         for _ in range(limiter.burst_size):
             allowed, _ = await limiter.is_allowed("user:exhaust")
+            assert allowed is True
         # Next request should be denied
         allowed, metadata = await limiter.is_allowed("user:exhaust")
         assert allowed is False