fix: change return type of nvml_to_cuda_map to int for consistency

[xrwang8]

Summary

• change nvml_to_cuda_map to return int so it can safely signal failure with -1
• update the function declaration and callers to rely on the signed return value

This removes the implementation-defined behavior where an unsigned -1 could be interpreted as a huge positive index, bypassing the <0 guard and leading to out-of-bounds access when NVML devices are not
visible to CUDA (e.g., under CUDA_VISIBLE_DEVICES remapping).

[hami-robot]

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by: xrwang8
Once this PR has been reviewed and has the lgtm label, please assign archlitchi for approval. For more information see the Kubernetes Code Review Process.

The full list of commands accepted by this bot can be found here.

Needs approval from an approver in each of these files:

• OWNERS

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment