test(mysql2): Harden tracing-channel healthcheck to fix flaky CI

The suite gated readiness on `mysqladmin ping`, which reports "alive"
even during MySQL 8's init bootstrap (it treats access-denied as alive).
So `docker compose up --wait` could return before the real server and
root password were ready, and the scenario's first connection was
dropped ("server closed the connection") — crashing the run and timing
out the test. Gate on an authenticated `SELECT 1` instead, mirroring the
knex/mysql2 sibling suite (#21868).

Author: logaretm

dev-packages/node-integration-tests/suites/tracing/mysql2-tracing-channel/docker-compose.yml

@@ -8,7 +8,11 @@ services:
     environment:
       MYSQL_ROOT_PASSWORD: password
     healthcheck:
-      test: ['CMD-SHELL', 'mysqladmin ping -h 127.0.0.1 -uroot -ppassword']
+      # `mysqladmin ping` reports "alive" even during MySQL's init bootstrap (it treats
+      # access-denied as alive), so `--wait` can return before the real server and root
+      # password are ready, and early connections get dropped ("server closed the
+      # connection"). Run an authenticated query instead so readiness gates on the real server.
+      test: ['CMD-SHELL', 'mysql -h 127.0.0.1 -uroot -ppassword -e "SELECT 1"']
       interval: 2s
       timeout: 3s
       retries: 30