update weather_forecast demo to spin up grafana and prometheus when using monitoring profile

has full dashboard with total requests, time per output token, time to
first token, total latency, output sequence length, and input sequence
length.

crates/llm_gateway/tests/integration.rs

@@ -290,7 +290,7 @@ fn llm_gateway_request_ratelimited() {
     let args = tester::MockSettings {
         wasm_path: wasm_module(),
         quiet: false,
-        allow_unexpected: true,
+        allow_unexpected: false,
     };
     let mut module = tester::mock(args).unwrap();