v0.47.0

🎯T71 — mnemo_compactor_status stops crying wolf.

Fixed

• "Watcher appears stuck" warning no longer fires when the watcher is actively ticking through a batch (🎯T71). Observed against v0.46.0's 7.7K compactor backlog: one scan returned 100 candidates and the watcher spent ~10 minutes ticking through 20 of them per the per-scan cap before the next scan ran, so last_scan_at legitimately sat well past 2 × scan_interval even though work was happening. The heuristic now treats last_tick_at as proof of life alongside last_scan_at — warns only when both are stale. Genuine wedged cases (no scans AND no ticks) still surface; startup before the first tick also still surfaces.