Fix OSRM healthcheck: use curl HTTP check and add start_period

/dev/tcp is bash-only and fails in the container's default sh.
Switch to a real HTTP check against the OSRM API root, and add a
30s start_period so Docker doesn't fail the check before the road
data finishes loading.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

docker-compose.yml

@@ -9,10 +9,11 @@ services:
       - ./docker/osrm/data:/data
     command: osrm-routed --algorithm mld /data/connecticut-latest.osrm
     healthcheck:
-      test: ["CMD-SHELL", "echo > /dev/tcp/localhost/5000"]
+      test: ["CMD-SHELL", "curl -sf http://localhost:5000/ || exit 1"]
-      interval: 30s
+      interval: 15s
       timeout: 5s
-      retries: 3
+      retries: 5
+      start_period: 30s
 
   balloons-shop:
     build: .