llm: split usage logs by agent type into subdirectories

llm-logs/fact-mine/2026-03-05.md, llm-logs/consolidate/2026-03-05.md,
etc. Makes it easy to review one agent at a time when debugging and
optimizing prompts.

src/llm.rs

@@ -12,7 +12,7 @@ use std::process::Command;
 
 fn log_usage(agent: &str, model: &str, prompt: &str, response: &str,
              duration_ms: u128, ok: bool) {
-    let dir = crate::config::get().data_dir.join("llm-logs");
+    let dir = crate::config::get().data_dir.join("llm-logs").join(agent);
     let _ = fs::create_dir_all(&dir);
 
     let date = chrono::Local::now().format("%Y-%m-%d");