Support for multiple NVIDIA GPUs

I'm encountering an issue when attempting to use the sysbox runtime with containers that require NVIDIA GPU access on a system with multiple GPUs. While the setup works seamlessly on a single-GPU machine, it fails when deployed on a multiple GPU machine.

The container should have access to all or specific GPUs as defined in the Docker Compose file, with GPU devices and drivers properly passed through by the sysbox runtime.

When deploying a container on the multi-GPU system, the following error occurs:

Failed to deploy a stack: compose up operation failed: Error response from daemon: failed to create task for container: failed to create shim task: OCI runtime create failed: container_linux.go:439: starting container process caused: process_linux.go:608: container init caused: Running hook #0:: error running hook: exit status 1, stdout: , stderr: Auto-detected mode as 'legacy' nvidia-container-cli: mount error: mount operation failed: /var/lib/docker/overlay2/e5409caee5c762014641d9a3fa7981fc960b3c2309980dda0e6b5d87b096a649/merged/proc/driver/nvidia: no such file or directory: unknown

Start by reproducing the Docker Compose deployment on a multi-GPU host and inspect how the sysbox runtime handles NVIDIA device and driver passthrough. Compare access to all GPUs with access to specific GPUs defined in Compose. Done means containers start successfully with the requested GPU devices and drivers.