fix(routing): address review feedback on CJK estimation and observability

1. CJK token estimation: replace flat rune_count/3 with script-aware
   counting — CJK runes (U+2E80–U+9FFF, U+F900–U+FAFF, U+AC00–U+D7AF)
   count as 1 token each, non-CJK runes at /4. This fixes a 3x
   underestimate for Chinese/Japanese/Korean text that could incorrectly
   route complex CJK messages to the light model.

2. Routing observability: SelectModel now returns the computed score as
   a third value. selectCandidates logs the score on both paths — Info
   level for light model selection, Debug level for primary model
   selection.

3. Added tests: TestExtractFeatures_TokenEstimate_Mixed (CJK+ASCII mix),
   TestRouter_SelectModel_ReturnsScore.

Addresses review feedback from @mingmxren.

pkg/agent/loop.go

@@ -1192,8 +1192,14 @@ func (al *AgentLoop) selectCandidates(
 		return agent.Candidates, agent.Model
 	}
 
-	_, usedLight := agent.Router.SelectModel(userMsg, history, agent.Model)
+	_, usedLight, score := agent.Router.SelectModel(userMsg, history, agent.Model)
 	if !usedLight {
+		logger.DebugCF("agent", "Model routing: primary model selected",
+			map[string]any{
+				"agent_id":  agent.ID,
+				"score":     score,
+				"threshold": agent.Router.Threshold(),
+			})
 		return agent.Candidates, agent.Model
 	}
 
@@ -1201,6 +1207,7 @@ func (al *AgentLoop) selectCandidates(
 		map[string]any{
 			"agent_id":    agent.ID,
 			"light_model": agent.Router.LightModel(),
+			"score":       score,
 			"threshold":   agent.Router.Threshold(),
 		})
 	return agent.LightCandidates, agent.Router.LightModel()