Causal diagnostic landing preparation

The separate story/combat-causal-landing branch rebases the nine diagnostic commits through 19aae6cd onto main 9cf06ec. The original branch and sealed rev2 products remain unchanged. Native05 passed all 60 tests on the original source; that evidence does not qualify this rebased tree. A fresh native build and both independent reviews remain required before landing.

Integration decisions:

• CMakeLists.txt merged automatically. Keep all main targets and its dynamic CTest registration, adding the diagnostic test target without restoring the old test inventory or hardcoding the previous count of 60.
• WorldPump.hpp merged automatically. Keep protocol 15/16 routing and the PopulationCut send/log branch. Run the optional noexcept diagnostic observer after that branch, preserving main's send results and statistics.
• EnemySync.cpp required manual resolutions. Keep the complete record population capture/resolution functions before the separate causal helpers. Keep main's record-authority HP log/readback and then run the optional causal readback. Keep PopulationCut dispatch immediately after the authenticated scope gate; add the rejection observation before the gate's existing continue.
• EntityHook.cpp merged automatically. Keep main's Limit admission install, ready checks, reinitialization refusals and shutdown ownership. Only add the existing opt-in diagnostic QPC configuration and enable hit tracing when that diagnostic is requested; no old Limit hooks are restored.

The diagnostic engine, NativeHitTrace implementation/header and three diagnostic control files match the original source byte-for-byte. Claim, HP and death wire tags/layouts used by the diagnostic are unchanged by protocol 15/16; main adds manifest record identity and PopulationCut independently. No deliberate change to causal admission, qualification, bounded raw receipts, loss retirement or default-off behavior is intended. Main's newer population and record-authority work can change when observations run; retained byte equality is not a proof that the combined implementation behaves identically. In particular, main's record-authority logging precedes the diagnostic post-store readback.

Offline checks: clean original branch still at 19aae6cd, main ancestry, range-diff inspection, unchanged diagnostic/control blobs, unchanged main protocol/codec and Limit admission blobs, preserved main record-population functions, and git diff --check. No native commands, push, PC2 contact or live actions were performed for this landing preparation.