Avoid full ledger reloads in operator snapshots

[yvette-carlisle]

Summary

• query Linear execution ledger rows by service/issue directly from SQLite for operator history hydration
• avoid refreshing the full runtime state cache on every history lane snapshot read
• add a cross-store regression proving issue-scoped ledger listing does not materialize the full ledger cache

Verification

• cargo test -p decodex persistent_linear_execution_event_listing_does_not_refresh_full_ledger --all-features
• cargo test -p decodex state_store_open_persists_runtime_history_across_instances --all-features
• cargo make check