fix(drivers): Downgrade amd-gpu driver deployment to version 1.18.0

Due to a currently unsolved issue, the device plugin reports 0 available
AMD GPUs on system that have fully functional GPUs ready to go. This
change in behaviour appeared when upgrading from Kubernetes 1.24.11 to
1.25.8 and is currently unclear.

References:
https://github.com/RadeonOpenCompute/k8s-device-plugin/issues/31