fix(distill): max_tokens 600→800 — Agnes推理模型JSON不再截断

engine.go: max_tokens 600→800
consolidate.go: max_tokens 300→800
根因: Agnes 2.0-flash是推理模型，max_tokens=600里大量是推理token，
正文只剩~300→JSON截断→parse error→蒸馏退化。
800确保推理+正文都够用。

go/internal/distill/consolidate.go

@@ -184,7 +184,7 @@ func (e *Engine) ConsolidateMemory(target MemoryCandidate, candidates []MemoryCa
 		"model":       e.LLMModel,
 		"messages":    []map[string]string{{"role": "user", "content": prompt}},
 		"temperature": 0.1,
-		"max_tokens":  300,
+		"max_tokens":  800,
 	}
 	jsonBody, err := json.Marshal(body)
 	if err != nil {

go/internal/distill/engine.go

@@ -339,7 +339,7 @@ func (e *Engine) callLLM5D(content string) (LLMResponse, error) {
 			{"role": "user", "content": prompt},
 		},
 		"temperature": 0.2,
-		"max_tokens":  600,
+		"max_tokens":  800,
 	}
 
 	jsonBody, err := json.Marshal(body)