Implement GPU endpoints with provider discovery about cloudmos HOT 2 CLOSED

PR #109 adds a /internal/gpu endpoint using this spec. Currently waiting on more gpu providers to upgrade before merging to make sure the returned data is accurate. The provider query param will accept either a provider address or a host uri.

This endpoint is relying on the assumption that all GPUs within a node are the same. If that ever changes, the provider's status endpoint should be changed accordingly for this endpoint to continue reporting accurate information.

from cloudmos.

the api response should be as following

{
  "gpus" : {
    "total": {
      "allocated": "1",
      "allocatable": "1",
    },
    "details": {
      "nvidia": [
        {
          "model": "a100",
          "ram": "40Gi",
          "interface": "PCIe|SXM4"
          "allocatable": 1,
          "allocated": 1
        }
      ]
    }
  }
}

• default query should return totals and details for all of the providers on the network
• query param provider filters results per provider
• query param vendor filters results per vendor
• query param model filters results per model
• query param memory_size filters results results with memory size

all query params above can be combined

from cloudmos.