add session pinning to llm_chat handler and rewrite session pinning demo

- extend brightstaff llm_chat_inner to extract X-Session-Id, check the
  session cache before routing, and cache the result afterward — same
  pattern as routing_service.rs
- replace old urllib-based demo with a real FastAPI research agent that
  runs 3 independent tool-calling tasks with alternating intents so
  Plano routes to different models; demo.py is a pure httpx client that
  shows the routing trace side-by-side with and without session pinning

demos/llm_routing/session_pinning/start_agents.sh

@@ -0,0 +1,28 @@
+#!/bin/bash
+set -e
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+PIDS=()
+
+log() { echo "$(date '+%F %T') - $*"; }
+
+cleanup() {
+    log "Stopping agents..."
+    for PID in "${PIDS[@]}"; do
+        kill "$PID" 2>/dev/null && log "Stopped process $PID"
+    done
+    exit 0
+}
+
+trap cleanup EXIT INT TERM
+
+export PLANO_URL="${PLANO_URL:-http://localhost:12000}"
+export AGENT_PORT="${AGENT_PORT:-8000}"
+
+log "Starting research_agent on port $AGENT_PORT..."
+uv run "$SCRIPT_DIR/agent.py" &
+PIDS+=($!)
+
+for PID in "${PIDS[@]}"; do
+    wait "$PID"
+done