BUG: recover_sub_pool no longer raises on concurrent create_actor

[AmirF194]

What do these changes do?

MainActorPool.recover_sub_pool (indigen/pool.py) iterates
self._allocated_actors[address].values() directly and awaits
self.call(...) inside the loop for every actor being recreated.
create_actor (pool.py) writes into that same dict across an await
too: it sets a placeholder key before calling out, then replaces or
pops it once the call returns. If a create_actor call lands on the
same sub pool address while recover_sub_pool is mid-loop, the dict's
size changes between two iteration steps and Python raises
RuntimeError: dictionary changed size during iteration, aborting
recovery for every actor still left in the loop.

The fix snapshots .values() into a list before iterating, so later
mutations of the live dict don't affect the iteration in progress.
monitor_sub_pools already guards its own iteration over
self.sub_processes the same way a few lines down, so this brings
recover_sub_pool in line with that existing pattern rather than
introducing a new one.

Related issue number

Fixes #48

Check code requirements

• tests added / passed
• Ensure all linting tests pass

Verification

• New test test_recover_sub_pool_concurrent_create_actor drives the
  real recover_sub_pool/create_actor methods on a MainActorPool
  instance with call/start_sub_pool/wait_sub_pools_ready stubbed,
  and controls the interleaving with asyncio.Events so a
  create_actor call is guaranteed to land while recovery is paused
  mid-iteration. Confirmed it fails on unmodified main with the
  exact RuntimeError from the issue, and passes on this branch, both
  in the same container.
• Full xoscar/backends suite: 82 passed, 1 pre-existing failure
  (test_sub_pool_quit_with_main_pool, a zombie-reaping assertion that
  fails identically on unmodified main in this container due to no
  init process; unrelated to this change) plus 6 skipped, both before
  and after the fix.
• black, flake8, isort and mypy (this repo's own pre-commit
  hooks) all clean on the two changed files.
• Not run: the real ci.yml matrix across all five Python versions
  and three OSes; verified on Python 3.11 in a Debian container only.

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 75.85%. Comparing base (d6465c9) to head (c8bf238).
⚠️ Report is 91 commits behind head on main.

Additional details and impacted files

@@             Coverage Diff             @@
##             main     #204       +/-   ##
===========================================
- Coverage   88.97%   75.85%   -13.13%     
===========================================
  Files          48       57        +9     
  Lines        4038     5326     +1288     
  Branches      770      581      -189     
===========================================
+ Hits         3593     4040      +447     
- Misses        358     1145      +787     
- Partials       87      141       +54     

Flag	Coverage Δ
unittests	75.42% <ø> (-13.39%)	⬇️

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.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[qinxuye]

LGTM

[mergify]

Tick the box to add this pull request to the merge queue (same as @mergifyio queue).

• Queue this pull request

[AmirF194]

Thanks for catching the placeholder replay case and for the quick review, appreciate it.