feat: reattach named child runs after migration or resurrection

[vdusek]

Actor.start, Actor.call and Actor.call_task accept an optional run_name. The SDK writes the name, the Actor or task ID and the run ID to the default key-value store (APIFY_CHILD_RUNS) right after the child starts. After a migration or resurrection of the parent, the same call finds the recorded run and reuses it:

• READY / RUNNING: reattaches to it
• SUCCEEDED: returns it as is
• ABORTED / TIMED-OUT: resurrects it (an ABORTING / TIMING-OUT run is waited for first)
• FAILED, or the run no longer exists: starts a new run and keeps the old ID in previousRunIds

A named call on a reattached or resurrected run streams only new log lines. Reusing a name for a different Actor or task raises ValueError, and concurrent calls under one name in the same process share a single run.

A hard kill between the platform starting the child and the registry write can still orphan the child. Closing that gap needs an idempotency key on the run-start endpoint.

Closes: #1127

✍️ Drafted by Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.12%. Comparing base (32a0a30) to head (ed8160c).

Additional details and impacted files

@@             Coverage Diff             @@
##           master    #1149       +/-   ##
===========================================
+ Coverage   53.03%   93.12%   +40.09%     
===========================================
  Files          53       54        +1     
  Lines        3560     3665      +105     
===========================================
+ Hits         1888     3413     +1525     
+ Misses       1672      252     -1420     

Flag	Coverage Δ
e2e	32.82% <39.44%> (+0.21%)	⬆️
integration	52.68% <39.44%> (-0.35%)	⬇️
unit	89.82% <100.00%> (?)

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.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[vdusek]

TODO:

• Rename the registry key from APIFY_CHILD_RUNS to a __-prefixed name, which is how other internal records in the default KVS are named (e.g. __STORAGE_ALIASES_MAPPING). The name isn't final yet (__NAMED_CHILD_RUNS, __CHILD_RUNS_STATE, ...) and should match the JS SDK (feat: named child runs that survive migration and resurrection apify-sdk-js#759). The change covers CHILD_RUNS_KEY, the docs in docs: describe how to compose Actors #1160 and this PR's description.
• Resurrect a recorded ABORTED / TIMED-OUT run with Actor.resurrect (feat: add Actor.resurrect to restart a finished Actor run with its original storages #1165, merged) in _find_or_start_child_run, which calls run_client.resurrect directly now.
• Forward max_items (fix: add missing max_items to Actor.start, Actor.call and Actor.call_task #1166, merged) in the named branches of Actor.start, Actor.call and Actor.call_task, both to the started run and to a resurrected one.

Both need master merged into the stack first.

✍️ Drafted by Claude Code

[Mantisus]

_wait_for_child_run waits on client.run(...) through wait_for_finish(), which tolerates 404s for a few seconds after a run starts. Here we use a bare get(), and a 404 from it starts a new run. Is that on purpose?

[Mantisus]

Can't we use only one for both operations? 🙂

[Mantisus]

We can use WeakValueDictionary here