training: per-node scoring with graph weight updates

Memory scoring now uses the graph as source of truth:
- last_scored timestamp on each node (new capnp field @22)
- Nodes scored when older than scoring_interval_secs (default 1hr)
- Oldest-scored-first ordering
- Window: scoring_response_window assistant responses (default 100)
- First-quarter memories scored even without full window
- Per-response normalization (raw divergence / response count)
- Asymmetric weight update: alpha=0.5 up, alpha=0.1 down
  (responds fast to importance, decays slowly — memories stay
  surfaced even if only useful 1/4 of the time)

Graph writes disabled pending normalization calibration.

Also: configurable scoring_interval_secs and scoring_response_window.

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

src/subconscious/subconscious.rs

@@ -104,14 +104,10 @@ pub struct AgentCycleState {
     log_file: Option<File>,
     pub agents: Vec<AgentInfo>,
     pub last_output: AgentCycleOutput,
-    /// Incremental memory scoring — entry index to resume from.
-    pub memory_score_cursor: usize,
     /// Whether incremental memory scoring is currently running.
     pub memory_scoring_in_flight: bool,
     /// Latest per-memory scores from incremental scoring.
     pub memory_scores: Vec<(String, f64)>,
-    /// Total unique memories in the context (updated when scoring starts).
-    pub memory_total: usize,
 }
 
 const AGENT_CYCLE_NAMES: &[&str] = &["surface-observe", "journal", "reflect"];
@@ -138,10 +134,8 @@ impl AgentCycleState {
                 reflection: None,
                 sleep_secs: None,
             },
-            memory_score_cursor: 0,
             memory_scoring_in_flight: false,
             memory_scores: Vec::new(),
-            memory_total: 0,
         }
     }
 
@@ -192,11 +186,11 @@ impl AgentCycleState {
             name: "memory-scoring".to_string(),
             pid: None,
             phase: if self.memory_scoring_in_flight {
-                Some(format!("scoring {}/{}", self.memory_scores.len(), self.memory_total))
+                Some("scoring...".into())
             } else if self.memory_scores.is_empty() {
                 None
             } else {
-                Some(format!("{}/{} scored", self.memory_scores.len(), self.memory_total))
+                Some(format!("{} scored", self.memory_scores.len()))
             },
             log_path: None,
         });