feat: add deployment of nvidia nim

[laurentgrangeau]

Adding deployment of NVIDIA NIMs on the ML platform

[arueth]

For the repo, we try and follow the Terraform on Google Cloud: Best practices for general style and structure, could you take a look at it and update the PR. Mainly around the Adopt a naming convention and using underscore to delimit multiple words.

Although not required, it is good practice to break out standard attributes and block attributes as well as alphabetize attributes in resources and data sources to make merges and updates from multiple people cleaner.

Also please ensure that you run terraform fmt on the code as this will be enforce when merging to main.

[arueth]

Also see my prior comment.

[arueth]

If this isn't going to be populated with a value, can you remove it?

[arueth]

Is HDML required or could another StorageClass also be used?

[arueth]

Could you also include a versions.tf that includes the pinned provider versions.

[arueth]

You will also need to add the license header to your .tf files.

[roberthbailey]

How different is this from the nim guide that we merged into https://github.com/GoogleCloudPlatform/ai-on-gke/tree/main/tutorials-and-examples/nvidia-nim?

[arueth]

Looks to be similar, except done with Terraform. We are consuming this in the ml-platform, so it will be referenced internally primarily.