Allow worker_pool_start_method='spawn' with the asynchronous prefork pool

[GangEunzzang]

Summary

Follow-up to #10339, as discussed there. Related: #6036 (spawn support has been requested there since 2020; with this change it is available for both prefork pool implementations).

#10339 rejects worker_pool_start_method='spawn' whenever the worker uses an event loop, on the assumption that the asynchronous prefork pool cannot work with spawned children. In practice it does: I ran AsynPool with spawned children on Linux and macOS with hard and soft time limits, worker replacement after a hard time limit, --autoscale, --max-tasks-per-child, acks_late + task_reject_on_worker_lost, revoke(terminate=True), pool_restart, grow/shrink, chains/groups/chords, ETA/retry, large payloads and warm/cold shutdown, without a difference to fork (details in #10339 (comment)). The wiring the PR description worried about — the child reporting inqW_fd back to the parent — holds up because spawnv_passfds preserves fd numbers.

The check also had a side effect: it keyed off w.use_eventloop rather than the pool class, so -P solo and -P threads with an async transport were rejected too.

This PR:

• removes the rejection in Pool.create(), so spawn works with both prefork pool implementations (and no longer blocks solo/threads);
• updates the CLI help and the docs accordingly, and notes in the setting docs that replacement children have to import the application within worker_proc_alive_timeout;
• adds smoke tests that run the worker with worker_pool_start_method = "spawn" against the smoke brokers (AsynPool): tasks run in SpawnPoolWorker-* children, and the pool keeps working after a hard time limit kill.

Testing

• Unit tests updated (test_create_spawn_with_async_pool now asserts the pool is created with forking_enable=False).
• New smoke tests pass locally.
• Manual: redis broker with --pool-start-method spawn at -c 2, hard time limit replacement and 50 tasks; -P solo --pool-start-method spawn now starts.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 88.95%. Comparing base (afeabe2) to head (cb0e646).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #10584      +/-   ##
==========================================
+ Coverage   88.91%   88.95%   +0.03%     
==========================================
  Files         153      153              
  Lines       20106    20104       -2     
  Branches     2364     2363       -1     
==========================================
+ Hits        17877    17883       +6     
+ Misses       1923     1919       -4     
+ Partials      306      302       -4     

Flag	Coverage Δ
unittests	88.93% <ø> (+0.02%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.