neuro: split into scoring, prompts, and rewrite modules

neuro.rs was 1164 lines wearing three hats:
- scoring.rs (401 lines): pure analysis — priority, replay queues,
  interference detection, consolidation planning
- prompts.rs (396 lines): agent prompt generation and formatting
- rewrite.rs (363 lines): graph topology mutations — hub
  differentiation, triangle closure, orphan linking

The split follows safety profiles: scoring never mutates, prompts
only reads, rewrite takes &mut Store. All public API re-exported
from neuro/mod.rs so callers don't change.

src/neuro/mod.rs

@@ -0,0 +1,29 @@
+// Neuroscience-inspired memory algorithms, split by concern:
+//
+//   scoring  — pure analysis: priority, replay queues, interference, plans
+//   prompts  — agent prompt generation and formatting
+//   rewrite  — graph topology mutations: differentiation, closure, linking
+
+mod scoring;
+mod prompts;
+mod rewrite;
+
+// Re-export public API so `neuro::` paths continue to work.
+
+pub use scoring::{
+    replay_queue, detect_interference,
+    consolidation_plan, format_plan,
+    daily_check,
+};
+
+pub use prompts::{
+    load_prompt,
+    consolidation_batch, agent_prompt,
+};
+
+pub use rewrite::{
+    refine_target, LinkMove,
+    differentiate_hub,
+    apply_differentiation, find_differentiable_hubs,
+    triangle_close, link_orphans,
+};