Implement metrics API

[r4victor]

Closes #1809

This PR:

• Implements runner metrics API. The runner collects cpu and memory metrics via cgroups fs and nvidia gpu metrics via nvidia-smi.
• Adds server background tasks to collect and delete job metrics. By default, metrics are collected every 10 seconds and live for 1 hour.
• Implements server metrics API.
• Implements dstack stats command to view run metrics.

Backward compatibility:

• The server handles old runners that don't provide metrics API and emits warnings that metrics collection fails.

Tested:

• Metrics of a run with cgroups v1 host.
• Metrics of a run with cgroups v2 host.
• Metrics of a run with no GPUs.
• Metrics of a run with one GPU.
• Metrics of a run with multiple GPUs.
• Metrics of a run executed by the old runner.
• Multi-replica server metrics collection.

TODOs:

• Support AMD GPU metrics.

dstack stats example:

✗ dstack stats hot-frog-1 -w
 NAME        CPU  MEMORY           GPU                        
 hot-frog-1  2%   15307MB/49152MB  #0 22764MB/24576MB 0% Util 

Metrics API example: GET /api/project/{project_name}/metrics/job/{run_name}

{
	"metrics": [
		{
			"name": "cpu_usage_percent",
			"timestamps": [
				"2024-10-14T08:27:59.721935+00:00"
			],
			"values": [
				147
			]
		},
		{
			"name": "memory_usage_bytes",
			"timestamps": [
				"2024-10-14T08:27:59.721935+00:00"
			],
			"values": [
				652775424
			]
		},
		{
			"name": "memory_working_set_bytes",
			"timestamps": [
				"2024-10-14T08:27:59.721935+00:00"
			],
			"values": [
				167317504
			]
		},
		{
			"name": "gpus_detected_num",
			"timestamps": [
				"2024-10-14T08:27:59.721935+00:00"
			],
			"values": [
				2
			]
		},
		{
			"name": "gpu_memory_usage_bytes_gpu0",
			"timestamps": [
				"2024-10-14T08:27:59.721935+00:00"
			],
			"values": [
				2097152
			]
		},
		{
			"name": "gpu_util_percent_gpu0",
			"timestamps": [
				"2024-10-14T08:27:59.721935+00:00"
			],
			"values": [
				0
			]
		},
		{
			"name": "gpu_memory_usage_bytes_gpu1",
			"timestamps": [
				"2024-10-14T08:27:59.721935+00:00"
			],
			"values": [
				2097152
			]
		},
		{
			"name": "gpu_util_percent_gpu1",
			"timestamps": [
				"2024-10-14T08:27:59.721935+00:00"
			],
			"values": [
				0
			]
		}
	]
}