Update nvidia DRA driver version to v25.3.0

Readme update for  A3 Mega

Update the blueprints using managed_lustre vars to 36T/500MB Tier

Refactoring in gke persistent module

Merge v1.67.0 into develop

Updated A3-mega and A4-high Slurm blueprints to adopt nvidia add repository scirpt.

update terraform-provider version to 7.3.0

Remove superfluous addition chs logs to cloud ops config

downloading  libnccl2 and libnccl-dev for a3u and a4h

Add nvidia-imex-* to list of held packages

When changing startup script content or changing partitions, following
error may occur:
╷
│ Error: Provider produced inconsistent final plan
│
│ When expanding the plan for module.slurm_controller.module.slurm_files.google_storage_bucket_object.nodeset_config["flexnodeset"] to include new values learned so far during apply, provider "registry.terraform.io/hashicorp/google" produced an invalid new value for .md5hash: was known, but now unknown.
│
│ This is a bug in the provider, which should be reported in the provider's own issue tracker.

With this change this is no longer the case

Fixing output for gke-storage module

Add mufaqam-gcl to cluster-toolkit-writers.json

Update GKE cluster and firewall rules module versions to use recent google provider

…add-filestore-pvc

Add Filestore, PV, and sample job template snippets to the GKE H4D blueprint

Add NUMA-aware scheduling in GKE clusters (enabled for G4)

Add slurm-gke blueprint

Migrate Kueue installation to use Helm chart

Update Toolkit release to v1.68.0

Updated the instance_image.family in a3ultra-vm.yaml to use ubuntu-accelerator-2204-amd64-with-nvidia-570 instead of nvidia-550 for improved compatibility and performance.

This reverts commit 4ff648b, reversing
changes made to 51bb92e.

This reverts commit 721983b, reversing
changes made to 4f8b592.

This pull request makes several updates to the GKE cluster and network modules, primarily focused on removing NUMA-aware scheduling support and aligning Terraform module and provider versions for improved compatibility. The changes simplify the configuration and ensure consistent dependency management across modules.

**Removal of NUMA-aware scheduling support:**
* Removed the `enable_numa_aware_scheduling` variable and all related configuration from the GKE cluster module, including the `kubelet_config` block and references in documentation and example files. [[1]](diffhunk://#diff-7939cd594b53ae6e59dae4629a32d7558e7c23123919d7b6e469ac18a57adddcL244-L259) [[2]](diffhunk://#diff-e54397224c9be21ab0ad72546e3d818fd2a4921bf593b2b6d7e881e6fc1d56e6L528-L533) [[3]](diffhunk://#diff-35b044e2245368feb59f14b7a63621200c0df5f4245b426552a09b8329705507L158) [[4]](diffhunk://#diff-e6090e2163c0286245ffc70056c158ea25acdeab329b5d21352fb007f80f4c73L125)

**Module and provider version alignment:**
* Updated the required versions for the `google` and `google-beta` Terraform providers from `>= 7.2` to `>= 6.16` in both the `versions.tf` and documentation files to standardize provider requirements. [[1]](diffhunk://#diff-b8e991c0f592027d61744d232494249832632ecc529153eac609f9e70444b471L21-R25) [[2]](diffhunk://#diff-35b044e2245368feb59f14b7a63621200c0df5f4245b426552a09b8329705507L106-R122)
* Changed the version constraints for the `workload_identity` module from `>= 40.0` to `~> 34.0` for compatibility, reflected in both code and documentation. [[1]](diffhunk://#diff-7939cd594b53ae6e59dae4629a32d7558e7c23123919d7b6e469ac18a57adddcL412-R396) [[2]](diffhunk://#diff-35b044e2245368feb59f14b7a63621200c0df5f4245b426552a09b8329705507L106-R122)
* Updated the version constraint for the `firewall_rule` module from `~> 12.0` to `~> 9.0` in both code and documentation for consistency. [[1]](diffhunk://#diff-bd07c7386bc0355d11578ce911bbc9a34a40f078b6f41fc0a8230d9b74eec28fL54-R54) [[2]](diffhunk://#diff-04a94d2869736107d8d67616c00f4e89cea5605aaec349c3d040b61be9cd1d0dL86-R86)
·

Updated the instance_image.family in a3ultra-vm.yaml to use ubuntu-accelerator-2204-amd64-with-nvidia-570 instead of nvidia-550 for improved compatibility and performance.