Register named remote actors and avoid late PgLeave respawns

[adamrobbie]

Summary

This PR makes named remote actors visible in the local named registry, matching the behavior of local actors. It also closes several lifecycle safety gaps that would have caused registry collisions and stale-message races.

Motivation

Addresses #93: "Remote actors don't appear in the local registry".

Remote actors already behave like local actors in most ways, but their names were not exposed through ractor::registry::where_is(...). That made name-based lookup inconsistent and reduced the usefulness of remote actors in cluster setups.

What changed

ractor/src/actor/actor_cell.rs

• Re-enabled name registration for remote actors in ActorCell::new_remote
• Name collisions (remote name already taken by a local actor) now log a warn! and skip registration instead of returning a hard SpawnErr. The shim is still spawned and reachable by pid.

ractor/src/actor/actor_ref.rs

• Documented that ActorRef::<T>::where_is returns None for remote actors because remote shims are typed as RemoteActorMessage internally. Callers should use registry::where_is directly for untyped remote lookups.

ractor_cluster/src/node/node_session.rs

• ActorFailed for a remote shim: kill and remove without immediately respawning. Respawning with the same pid reintroduces stale-message races and registry collisions; the remote node controls re-announcement via a new Spawn control message.
• PgLeave for an unknown pid: log a debug message and skip instead of recreating the shim. If the actor already terminated locally, group membership was already cleaned up by the shutdown path.

PID registry behavior unchanged:

• Local actors still register in the PID registry
• Remote actors do not register in the PID registry

Regression tests added

• Named remote actor appears in registry::where_is(...) and is removed on shutdown
• Remote spawn with a colliding local name succeeds without evicting the local actor
• ActorFailed on a remote shim does not respawn it
• Late PgLeave after remote actor termination does not respawn the shim

Testing

• cargo fmt --all
• cargo clippy --all -- -D clippy::all -D warnings
• cargo test -p ractor --features cluster remote_actor_with_name_is_registered -- --nocapture
• cargo test -p ractor --features cluster remote_actor_name_collision_with_local_does_not_fail_spawn -- --nocapture
• cargo test -p ractor_cluster node_session_handle_control -- --nocapture
• cargo test -p ractor_cluster actor_failed_does_not_respawn_remote_actor -- --nocapture
• cargo test -p ractor_cluster pg_leave_after_terminate_does_not_respawn_remote_actor -- --nocapture

[adamrobbie]

And thank you for a library like this.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 86.22%. Comparing base (4518baa) to head (6039f09).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #421      +/-   ##
==========================================
+ Coverage   86.00%   86.22%   +0.21%     
==========================================
  Files          73       73              
  Lines       13564    13791     +227     
==========================================
+ Hits        11666    11891     +225     
- Misses       1898     1900       +2     

☔ 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.

[slawlor]

can you fix the failing tests? Then I'm happy to review.