test: xfail a flaky app test in CI (#9361)

Many of the app composition and embedding tests are flaky in CI. This
suggests either cross-test state pollution or a bug in the underlying
implementation.

These bugs are not high priority to fix, so the flaky test has been
marked as xfail. If this ever does become
a priority (in particular, if we see GitHub issues about these methods
failing in real usage), these failing tests may provide a clue.

This test appears to only have started failing somewhat consistently
after #9257, but I don't see any relation between the two (it's possible
#9257's tests may pollute the test environment, but I haven't been able
to see how).


<!-- This is an auto-generated description by cubic. -->
---
## Summary by cubic
Marked
tests/_ast/test_app.py::TestAppComposition::test_overrides_returns_multiple_defs
as xfail to stabilize CI for app composition/embedding. The failure is
CI-only and likely due to cross-test state or an underlying bug; we’ll
revisit if it affects users.

<sup>Written for commit 3046298.
Summary will update on new commits.</sup>

<!-- End of auto-generated description by cubic. -->

Author: akshayka

tests/_ast/test_app.py

@@ -1117,6 +1117,12 @@ def __():
     assert "value2" in output
 
 
+# TODO(akshayka): Many of the app composition and embedding tests are flaky in
+# CI. This suggests either cross-test state pollution or a bug in the
+# underlying implementation. These bugs are not high priority to fix, so
+# the flaky tests have been marked as xfail. If this ever does become
+# a priority (in particular, if we see GitHub issues about these methods
+# failing in real usage), these failing tests may provide a clue.
 class TestAppComposition:
     async def test_app_embed(self) -> None:
         app = App()
@@ -1784,6 +1790,10 @@ def __(x: int) -> tuple[int]:
         assert not k.errors
         assert k.globals["overrides"] == {"x": 100}
 
+
+    @pytest.mark.xfail(
+        True, reason="Flaky in CI, can't repro locally", strict=False
+    )
     async def test_overrides_returns_multiple_defs(
         self, k: Kernel, exec_req: ExecReqProvider
     ) -> None:
@@ -1827,8 +1837,6 @@ def __(a: int, b: int) -> tuple[int]:
         assert k.globals["overrides"] == {"a": 10, "b": 20}
 
 
-
-
 class TestAppKernelRunnerRegistry:
     def test_get_runner(self, k: Kernel) -> None:
         # `k` fixture installs a context, needed for AppKernelRunner