RDMA/mthca: Use generic resize-CQ lock Replace the open‑coded resize‑CQ lock with the standard core implementation for better consistency and maintainability. Signed-off-by: Leon Romanovsky <leonro@nvidia.com>