FleetIntelService getActors single-chunk pagination scaling gap

[NickCrew]

Description

FleetIntelService.getActors uses a single-chunk read with limit cap (~200 rows) where the legacy ingestion service paginated 100×10. With realistic fleet cardinalities this caps fleet-view actor results before dedup runs, which would silently truncate TASK-79 fleet routes and break tests at >200 distinct actors. Investigate moving to cursor-paged reads or raising the cap with batching, and add a guard that errors when the cap is hit (so we don't silently lose rows).

Acceptance Criteria

• Migrate synapse-waf to Rust 2024 edition #1 Read path for SensorIntelActor returns all rows above a single-page cap (cursor pagination or batched chunks), no silent truncation
• Cache parsed EvalContext data across body-phase and deferred WAF passes #2 If a cap is enforced, the API surface returns a structured warning rather than dropping rows
• Instantiate and integrate HeaderProfiler for per-endpoint header anomaly detection #3 Test covers > cap distinct actors across the fleet

---

Migrated from backlog TASK-98 (synapse).