v3.10.15

Resolves the "four contradictory stat sources" item ADR-074 deferred to a future round.

The four sources turned out **not** to be duplicates — they authoritatively measure four different layers (globalStats = trajectory-pipeline counters, sonaCoordinator = in-process SONA, memory-bridge = AgentDB entries, neural-patterns = neural store rows). So the fix is to aggregate the *view*, not the store.

**New surface**
- `getUnifiedLearningStats()` — returns all 4 sub-views with each sub-view naming its source path
- `hooks_intelligence_unified-stats` MCP tool exposing it
- `getMemoryBridgeStats()` + `getNeuralStoreStats()` — exported helpers
- A `consistency` block that *flags* cross-store drift (e.g. globalStats reports N patterns but neural_patterns is empty) instead of silently disagreeing

**Verification**
- 7 cross-store consistency tests
- Benchmark §F observed: global=10/11 tracks SONA, bridge=10 rows, neural=10, sona.available=true, 1 consistency note correctly flagging the pretrain-vs-neural-store gap
- 123/123 across unified-stats + self-learning + mcp-tools-deep

**Install:** `npx ruflo@3.10.15`