Rewrite PG to fix race condition problems

[slawlor]

1. ractor/src/pg.rs — Full rewrite of PgState and all functions

New GroupState struct bundles members and per-group listeners into one atomically-accessible unit: struct GroupState { members: HashMap<ActorId, ActorCell>, listeners: Vec, }

New PgState uses three maps instead of the old three separate maps:

• map: DashMap<ScopeGroupKey, GroupState> — unified members + per-group listeners
• index: DashMap<ScopeName, HashSet> — secondary index (now uses HashSet instead of Vec)
• world_listeners: DashMap<ScopeGroupKey, Vec> — scope-level/global monitors only

Key fixes:

• Cross-map inconsistency (1): Members and per-group listeners are now in one entry, atomically updated
• Zombie actor insertion (3): join_scoped filters out actors with status > Draining
• Notifications under lock (4): All functions clone listeners and drop entry guards before sending notifications
• leave_all misses global monitors (5): New notify_world_listeners helper checks both scoped and ALL_SCOPES_NOTIFICATION keys
• demonitor_scope incomplete (6): Now removes from world_listeners directly (clean)
• monitor_scope duplicate notifications (7): Registers ONLY in world_listeners, not also per-group
• leave_all cleanup race (9): Re-checks emptiness under entry lock before removing

2. ractor/src/actor/actor_cell.rs — Guard double cleanup (8)

Added && self.get_status() < ActorStatus::Stopping condition so cleanup only runs once on the first transition to Stopping, avoiding redundant full-DashMap iterations.

3. ractor/src/pg/tests.rs — 3 new tests

• test_monitor_scope_no_duplicate_notifications — verifies exactly 1 notification when monitor_scope is called after group exists
• test_demonitor_scope_fully_unsubscribes — verifies no notifications after demonitor_scope
• test_stopped_actor_not_inserted — verifies stopped actors are rejected by join

Resolves #374

[codecov]

Codecov Report

❌ Patch coverage is 86.52174% with 31 lines in your changes missing coverage. Please review.
✅ Project coverage is 85.73%. Comparing base (e6174c8) to head (26c5fa6).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
ractor/src/pg/tests.rs	71.92%	16 Missing ⚠️
ractor/src/pg.rs	91.22%	15 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #411      +/-   ##
==========================================
+ Coverage   85.56%   85.73%   +0.16%     
==========================================
  Files          73       73              
  Lines       13329    13361      +32     
==========================================
+ Hits        11405    11455      +50     
+ Misses       1924     1906      -18     

☔ View full report in Codecov by Sentry.
📢 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.

[github-actions]

Benchmark Comparison (PR vs main)

Results

group                                                                   new
-----                                                                   ---
Creation of 100 actors                                                  1.00   443.1±29.58µs        ? ?/sec
Creation of 1000 actors                                                 1.00      3.9±0.11ms        ? ?/sec
Waiting on 100 actors to process first message                          1.00   104.5±16.16µs        ? ?/sec
Waiting on 100000 messages to be processed                              1.00     32.0±2.67ms        ? ?/sec
Waiting on 50 messages with large data in the Future to be processed    1.00     70.3±5.61µs        ? ?/sec
Waiting on 500 actors to process first message                          1.00   243.1±67.71µs        ? ?/sec
factory_queuer_dispatch/workers/10                                      1.00  1509.5±44.73µs 323.5 KElem/sec
factory_queuer_dispatch/workers/100                                     1.00  1511.0±58.04µs 323.2 KElem/sec
large_messages/256_byte_messages                                        1.00     66.5±2.90µs 1467.7 KElem/sec
large_messages/64_byte_messages                                         1.00     57.0±4.53µs 1712.8 KElem/sec
small_messages/16_byte_messages                                         1.00     54.9±4.35µs 1778.5 KElem/sec
small_messages/32_byte_messages                                         1.00     54.0±2.73µs 1808.4 KElem/sec
small_messages/8_byte_messages                                          1.00     56.5±3.23µs 1728.6 KElem/sec
spawn_throughput/10                                                     1.00     59.5±0.75µs 164.1 KElem/sec
spawn_throughput/100                                                    1.00    455.3±7.88µs 214.5 KElem/sec

Benchmarks run on shared CI hardware (ubuntu-latest). Differences under ~5% may be noise.