feat: add gpu info to health state [TAB-162]

[vodkaslime]

What this PR is about

Adding GPU information to health state if possible - when cuda is supported in the environment. In environments where cuda is not supported (e.g. in MacOS or in docker containers where --gpus is not specified), just return with empty data.

In AWS g4dn instances where gpus are specified, examples are like:

{
    "model": "TabbyML/T5P-220M",
    "device": "cuda",
    "compute_type": "auto",
    "arch": "x86_64",
    "cpu_info": "Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz",
    "cpu_count": 4,
    "gpu_info": [
        "Tesla T4"
    ],
    "version": {
        "build_date": "2023-08-20",
        "build_timestamp": "2023-08-20T18:16:17.790164841Z",
        "git_sha": "VERGEN_IDEMPOTENT_OUTPUT",
        "git_describe": "VERGEN_IDEMPOTENT_OUTPUT"
    }
}

In AWS g4dn instances where gpus are not specified, examples are like:

{
    "model": "TabbyML/T5P-220M",
    "device": "cpu",
    "compute_type": "auto",
    "arch": "x86_64",
    "cpu_info": "Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz",
    "cpu_count": 4,
    "gpu_info": [],
    "version": {
        "build_date": "2023-08-20",
        "build_timestamp": "2023-08-20T18:16:17.790164841Z",
        "git_sha": "VERGEN_IDEMPOTENT_OUTPUT",
        "git_describe": "VERGEN_IDEMPOTENT_OUTPUT"
    }
}

[wsxiaoys]

Otherwise LGTM