perf: fix model reload thrash and cut time-to-first-token

Ollama reloaded the chat model mid-message because plan/summary/router
calls sent divergent runner options (no num_ctx/num_batch) and omitted
keep_alive — the main cause of perceived slowness.

- share one keep-alive httpx.AsyncClient for all Ollama calls
- unify runner options (runner_options) + keep_alive on every model
  call, embeddings included
- drop the blocking LLM routing fallback (pure lexical heuristic)
- run RAG recall + plan + code-model pick in parallel inside the SSE
  stream, after the start event
- RAG cosine scoring off the event loop; cache /api/tags 30s and
  nvidia-smi 5s; frontend polls 2s->5s, warm poll backoff, dedup
  config fetch

feat: working session menu in TopBar (switch/create/rename/delete)
feat: workspace file deletion (DELETE /api/files + UI trash buttons)
docs: recommended Ollama env vars (KEEP_ALIVE, MAX_LOADED_MODELS...)

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

frontend/src/components/Icon.tsx

@@ -85,6 +85,12 @@ export const DownloadIcon = ({ size = 14, className }: Props) => (
   </svg>
 );
 
+export const TrashIcon = ({ size = 14, className }: Props) => (
+  <svg {...base(size, className)} strokeWidth={1.9}>
+    <path d="M4 7h16M10 7V5h4v2M9 7l.8 13h4.4L15 7M6.5 7l.7 13h9.6l.7-13" />
+  </svg>
+);
+
 export const CopyIcon = ({ size = 14, className }: Props) => (
   <svg {...base(size, className)} strokeWidth={1.7}>
     <rect x="9" y="9" width="11" height="11" rx="2" />