switch memory scoring to /v1/score endpoint

Replace prompt_logprobs-based scoring with the new vLLM /v1/score
endpoint. Much simpler: one API call per memory drop, returns
per-message total_logprob directly. No chunking needed, no OOM risk
— the endpoint only computes logits for scored tokens.

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

src/agent/types.rs

@@ -228,6 +228,15 @@ impl Message {
         self.content.as_ref().map_or("", |c| c.as_text())
     }
 
+    pub fn role_str(&self) -> &str {
+        match self.role {
+            Role::System => "system",
+            Role::User => "user",
+            Role::Assistant => "assistant",
+            Role::Tool => "tool",
+        }
+    }
+
     fn now() -> Option<String> {
         Some(Utc::now().to_rfc3339_opts(chrono::SecondsFormat::Secs, true))
     }