Adjust installation commands to NVIDIA suggestion

When you try to install CUDA drivers following our documentation, installation finishes with the following error using Centos-7.9 image with a NCv3 when you execute nvidia-smi

- "NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running."

After trying several times I went to NVIDIA documentation and using their steps works without any issue. Our article installs *cuda-drivers* package that has a depencendy with *nvidia-driver-latest-dkms*, while NVIDIA install first  *nvidia-driver-latest-dkms* and later *cuda-drivers*

Author: jangelfdez

articles/virtual-machines/linux/n-series-driver-setup.md

@@ -132,9 +132,10 @@ For example, CentOS 8 and RHEL 8 will need the following steps.
    sudo rpm -Uvh https://dl.fedoraproject.org/pub/epel/epel-release-latest-8.noarch.rpm
    sudo yum install dkms
    
-   wget https://developer.download.nvidia.com/compute/cuda/repos/rhel8/x86_64/cuda-rhel8.repo /etc/yum.repos.d/cuda-rhel8.repo
-
-   sudo yum install cuda-drivers
+   sudo yum-config-manager --add-repo https://developer.download.nvidia.com/compute/cuda/repos/rhel7/x86_64/cuda-rhel7.repo
+   sudo yum clean all
+   sudo yum -y install nvidia-driver-latest-dkms
+   sudo yum -y install cuda-drivers
    ```
 
 4. To optionally install the complete CUDA toolkit, type: