Add graph_topology, graph_health, interference_pairs tools

Convert {{topology}}, {{health}}, {{pairs}} placeholders to
{{tool:}} calls. Made format_topology_header, format_health_section,
format_pairs_section pub so tools can call them.

Co-Authored-By: Proof of Concept <poc@bcachefs.org>

src/subconscious/agents/challenger.agent

@@ -46,7 +46,7 @@ For each target node, one of:
 - **Don't be contrarian for its own sake.** If a node is correct,
   say so and move on.
 
-{{TOPOLOGY}}
+{{tool: graph_topology}}
 
 {{SIBLINGS}}
 

src/subconscious/agents/connector.agent

@@ -79,7 +79,7 @@ Set with: `poc-memory graph link-set <source> <target> <strength>`
 If you see default-strength links (0.10 or 0.30) in the neighborhoods
 you're exploring and you have context to judge them, reweight those too.
 
-{{TOPOLOGY}}
+{{tool: graph_topology}}
 
 ## Nodes to examine for cross-community connections
 

src/subconscious/agents/extractor.agent

@@ -44,7 +44,7 @@ pattern you've found.
 - **Preserve diversity.** Multiple perspectives on the same concept are
   valuable. Only delete actual duplicates.
 
-{{TOPOLOGY}}
+{{tool: graph_topology}}
 
 ## Neighborhood nodes
 

src/subconscious/agents/health.agent

@@ -36,8 +36,8 @@ overall structure.
 - Most output should be observations about system health. Act on structural
   problems you find — link orphans, refine outdated nodes.
 
-{{topology}}
+{{tool: graph_topology}}
 
 ## Current health data
 
-{{health}}
+{{tool: graph_health}}

src/subconscious/agents/replay.agent

@@ -40,7 +40,7 @@ clusters and determine how it fits.
 - **Trust the decay.** Unimportant nodes don't need pruning — just
   don't link them.
 
-{{TOPOLOGY}}
+{{tool: graph_topology}}
 
 ## Nodes to review
 

src/subconscious/agents/separator.agent

@@ -35,8 +35,8 @@ overlapping inputs and orthogonalize them.
 - **Session summaries are the biggest source of interference.**
 - **Look for the supersession pattern.**
 
-{{topology}}
+{{tool: graph_topology}}
 
 ## Interfering pairs to review
 
-{{pairs}}
+{{tool: interference_pairs}}

src/subconscious/agents/transfer.agent

@@ -45,7 +45,7 @@ entries, and extract those patterns into semantic nodes.
 - **The best extractions change how you think, not just what you know.**
   Extract the conceptual version, not just the factual one.
 
-{{TOPOLOGY}}
+{{tool: graph_topology}}
 
 {{SIBLINGS}}
 

src/subconscious/prompts.rs

@@ -23,7 +23,7 @@ pub struct AgentBatch {
     pub node_keys: Vec<String>,
 }
 
-pub(super) fn format_topology_header(graph: &Graph) -> String {
+pub fn format_topology_header(graph: &Graph) -> String {
     let sigma = graph.small_world_sigma();
     let alpha = graph.degree_power_law_exponent();
     let gini = graph.degree_gini();
@@ -139,7 +139,7 @@ pub(super) fn format_nodes_section(store: &Store, items: &[ReplayItem], graph: &
     out
 }
 
-pub(super) fn format_health_section(store: &Store, graph: &Graph) -> String {
+pub fn format_health_section(store: &Store, graph: &Graph) -> String {
     use crate::graph;
 
     let health = graph::health_report(graph, store);
@@ -195,7 +195,7 @@ pub(super) fn format_health_section(store: &Store, graph: &Graph) -> String {
     out
 }
 
-pub(super) fn format_pairs_section(
+pub fn format_pairs_section(
     pairs: &[(String, String, f32)],
     store: &Store,
     graph: &Graph,