Add TTS voice deletion, async speech jobs, voice-list cache, ref-seconds name tags

- DELETE /tts/voices/{name}: remove a cloned voice via the llama-swap
  passthrough (upstream chatterbox-tts-api exposes DELETE /voices/{name}).
- POST/GET/DELETE /tts/speech/jobs: durable job flow for long syntheses —
  dispatch returns 202 + job id, the synth queues on the GPU permit instead
  of fast-failing 429, and clients poll for the result (kept ~10 min).
- GET /tts/voices now serves an in-memory cache so listing voices doesn't
  make llama-swap spin up the TTS model (evicting the resident LLM);
  invalidated on create/delete, ?refresh=1 forces an upstream re-query.
- Created voice names are tagged with LLAMA_SWAP_TTS_REF_SECONDS (e.g.
  grandma-30s) so the library shows which ref length produced each clone.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

src/ai/llamacpp.rs

@@ -253,6 +253,34 @@ impl LlamaCppClient {
         resp.json().await.context("parsing create_voice response")
     }
 
+    /// Delete a cloned voice from the Chatterbox voice library
+    /// (`DELETE /voices/{name}` on the upstream, via llama-swap passthrough).
+    pub async fn delete_voice(&self, voice_name: &str) -> Result<Value> {
+        let url = format!(
+            "{}/upstream/{}/voices/{}",
+            self.swap_root(),
+            self.tts_model,
+            voice_name
+        );
+        let resp = self
+            .client
+            .delete(&url)
+            .send()
+            .await
+            .with_context(|| format!("DELETE {} failed", url))?;
+
+        if !resp.status().is_success() {
+            let status = resp.status();
+            let text = resp.text().await.unwrap_or_default();
+            bail!("llama-swap delete_voice failed: {} — {}", status, text);
+        }
+        // Some upstreams reply with an empty body on delete.
+        Ok(resp
+            .json()
+            .await
+            .unwrap_or_else(|_| json!({ "status": "deleted" })))
+    }
+
     /// Translate canonical messages to the OpenAI-compatible wire shape.
     /// Behaviorally identical to `OpenRouterClient::messages_to_openai` —
     /// stringify tool-call arguments, rewrite images into content-parts, attach