feat: add nvidia shared GPU settings to k8s model

[monirul]

Issue #, if available:

Description of changes:
This PR introduces two new fields within the NvidiaDevicePluginSettings structure: max_sharing_per_gpu and rename_shared_gpu. These new fields are for managing the timeslicing capabilities of Nvidia GPUs.

Setting max_sharing_per_gpu to 1 deactivates time-slicing, effectively disabling GPU sharing. Conversely, a value greater than 1 activates time-slicing, resulting in the creation of the specified number of GPU replicas.

The rename_shared_gpu field dictates the naming convention for shared GPU resources. A false value retains the original GPU name, while true triggers a renaming process, appending .shared to the GPU identifier. For instance, enabling this feature would transform nvidia.com/gpu to nvidia.com/gpu.shared.

By contributing this pull request, I grant permission to use, modify, replicate, and distribute my submission, in accordance with your preferred terms.

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice."

[cbgbt]

The code looks good to me! Do you mind rewriting the git commit message? I'll implement a linter for this, but the settings SDK tends to follow the conventional commit style.

You could rewrite the message as:

feat: add nvidia shared GPU settings to k8s model