feat(raft-log-replication): simulator AppendEntries delivery (task 8.2)

raft_sim_deliver_append_entries drains the AE queue, routes each buffered AE to
the (non-partitioned) destination via raft_on_append_entries_request, and routes
the result back to the source via raft_on_append_entries_reply — mirroring the
vote-RPC routing; partitioned endpoints are skipped.

Review-applied (HIGH): raft_sim_ae_t now captures the leader's SEND-time term
(leader_term) at buffer time and delivery uses it instead of the live source
state.term — without this, reorder/churn delivery (9.3/9.4) would inject the
wrong term and defeat UpdateTerm/stale-drop (G5). Smoke test asserts replication
to a connected follower + isolation of a partitioned one. ASAN green.

.kiro/specs/raft-log-replication/tasks.md

@@ -144,7 +144,8 @@
   - _Boundary: raft_sim.c_
   - _Depends: 6.3, 7.2_
 
-- [ ] 8.2 Add synchronous AppendEntries delivery honoring partitions
+- [x] 8.2 Add synchronous AppendEntries delivery honoring partitions
+  - NOTE (impl 2026-06-21): `raft_sim_ae_t` now carries `leader_term` captured at SEND time (in `sim_send_append_entries`) and delivery uses `ae->leader_term`, NOT the live source `state.term`. Required so reorder (9.3) and stale-term churn (9.4) deliver the sending-epoch term — using the live term would bypass UpdateTerm/stale-drop and defeat the G5 regression. Caught in review.
   - Add `raft_sim_deliver_append_entries`: for each buffered AE, call `raft_on_append_entries_request` on the (non-partitioned) destination, then route the reply back via `raft_on_append_entries_reply`; skip partitioned nodes (mirrors the existing vote-RPC routing)
   - Observable completion: driving one append + delivery replicates the entry to a connected follower; partitioned follower receives nothing
   - _Requirements: 8.1_