llm: full per-agent usage logging with prompts and responses

Log every model call to ~/.claude/memory/llm-logs/YYYY-MM-DD.md with
full prompt, response, agent type, model, duration, and status. One
file per day, markdown formatted for easy reading.

Agent types: fact-mine, experience-mine, consolidate, knowledge,
digest, enrich, audit. This gives visibility into what each agent
is doing and whether to adjust prompts or frequency.

src/fact_mine.rs

@@ -248,7 +248,7 @@ pub fn mine_transcript(path: &Path, dry_run: bool) -> Result<Vec<Fact>, String>
         eprint!("  Chunk {}/{} ({} chars)...", i + 1, chunks.len(), chunk.len());
 
         let prompt = format!("{}{}", prompt_prefix, chunk);
-        let response = match llm::call_haiku(&prompt) {
+        let response = match llm::call_haiku("fact-mine", &prompt) {
             Ok(r) => r,
             Err(e) => {
                 eprintln!(" error: {}", e);