Stabilize database workflow by retrying SQL Server PDO readiness

[Copilot]

The database GitHub Actions job was failing in run 88133959937 during SQL Server preparation. The workflow treated an open TCP port as readiness, but SQL Server was not always ready for authenticated PDO connections yet.

• Root cause

  • SQL Server startup race in CI: port 1433 becomes reachable before the server is ready to accept PDO login/query operations.

• Workflow changes

  • In .github/workflows/ci.yml (database job, Aguardar e preparar SQL Server step):
    • Kept the existing TCP port wait.
    • Replaced single-shot PDO admin connect + DB creation with a bounded retry loop.
    • Added explicit terminal error when PDO readiness does not happen within retry budget.

• Behavioral impact

  • Reduces flaky failures caused by transient SQL Server warm-up delays without changing test scope or DB semantics.

for attempt in {1..30}; do
  if php -r '$pdo = new PDO(getenv("SISP_TEST_SQLSRV_ADMIN_DSN"), getenv("SISP_TEST_SQLSRV_USER"), getenv("SISP_TEST_SQLSRV_PASSWORD")); if ((int) $pdo->query("SELECT DB_ID(NCHAR(115) + NCHAR(105) + NCHAR(115) + NCHAR(112))")->fetchColumn() === 0) { $pdo->exec("CREATE DATABASE [sisp]"); }'; then
    break
  fi
  if [ "$attempt" -eq 30 ]; then
    echo 'SQL Server abriu a porta 1433, mas não aceitou conexões PDO a tempo.' >&2
    exit 1
  fi
  sleep 2
done