Fix local_ssd_config issue that forces node-pool recreation

[sharabiani]

This PR fixes the issue that when ephemeral_storage_local_ssd_config is not set, it forces node-pool recreation on re-deploy.

Key Changes:

disk_definitions implemented:

It contain default values for ephemeral_storage_local_ssd_config and local_ssd_count_nvme_block per machine type. It's defined based on the values here.

Selection priority order:

1. variable local_ssd_count_ephemeral_storage and local_ssd_count_nvme_block if any is not null
2. disk_definitions if not empty for the machine_type
3. default to null value for both local_ssd_count_ephemeral_storage and local_ssd_count_nvme_block

Required changes to use the new local variable

Manual testing:

Scenario 1:

1. Deployed examples/gke-a3-highgpu.yaml
2. Run the deploy again.
3. Verified NO destruction and recreation of the node-pool is required.

Scenario 2:

1. Deployed examples/gke-a3-highgpu.yaml
2. Set local_ssd_count_ephemeral_storage to 16 explicitly
3. Run the deploy again.
4. Verified NO destruction and recreation of the node-pool is required.

Scenario 2:

1. Deployed examples/gke-a3-highgpu.yaml
2. Removed "a3-highgpu-8g" from disk_definitions
3. Run the deploy again.
4. Verified destruction and recreation of the node-pool is required.

Additional Notes:

• Made changes to relevant documentation.