Give macOS CI extra headroom for cancel-cascade tests

The `cpu_scaling_factor()` CI bump (2x) wasn't enough for the
macOS runners — slower + noisier than linux for our multi-actor
cancel-cascade timing — so `test_nested_multierrors[depth=3]` and
`test_fast_graceful_cancel_*` flaked there (linux + sdist green),

- make the CI headroom platform-aware: 3x on macOS, 2x on linux
  (keeps the proven linux budget; depth=3 inner 12*3=36s still
  fits under the 40s outer wall).
- give `test_fast_graceful_cancel_*` the `cpu_scaling_factor()`
  headroom it was missing entirely (was a bare `timeout=2.9`).

Verified locally w/ `CI=1` (2x linux path; 3x macOS path confirmed
via forced `_non_linux`).

(this patch was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code

tests/conftest.py

@@ -124,8 +124,12 @@ def cpu_scaling_factor() -> float:
     # headroom. Apply a flat CI bump so every timing-test deadline
     # /assert that keys off this factor gets headroom on CI HW
     # (compounds with any local-throttle factor).
+    #
+    # macOS runners are noticeably slower + noisier than the linux
+    # ones for our multi-actor cancel-cascade tests, so give them
+    # extra headroom (3x vs 2x).
     if _ci_env:
-        factor *= 2
+        factor *= 3 if _non_linux else 2
 
     return factor
 

tests/test_cancellation.py

@@ -955,6 +955,12 @@ def test_fast_graceful_cancel_when_spawn_task_in_soft_proc_wait_for_daemon(
     if _friggin_windows:  # smh
         timeout += 1
 
+    # CPU-scaling / CI latency headroom — macOS CI especially is
+    # slow for this graceful-vs-hard-reap timing race; see
+    # `cpu_scaling_factor()`.
+    from .conftest import cpu_scaling_factor
+    timeout *= cpu_scaling_factor()
+
     async def main():
         start = time.time()
         try: