AllWatcher does not report agents that are down

The JIMM service relies on the AllModelWatcher to observe changes in status of the model. One of the things it needs to know is when machines and units are down. Unfortunately the allwatcher does not provide this information, even though it's reported on a full status.

For example, here is the reply to a FullStatus API call where the machine is down:

 {
  request-id: 2
  response: {
   applications: {
    ubuntu: {
     can-upgrade-to: ""
     charm: "cs:ubuntu-12"
     endpoint-bindings: {}
     exposed: false
     life: ""
     meter-statuses: null
     public-address: ""
     relations: {}
     series: "xenial"
     status: {
      data: {}
      info: "ready"
      kind: ""
      life: ""
      since: "2018-08-23T14:16:31.26037557Z"
      status: "active"
      version: ""
     }
     subordinate-to: null
     units: {
      "ubuntu/2": {
       agent-status: {
        data: {}
        info: ""
        kind: ""
        life: ""
        since: "2018-08-23T14:16:32.792690804Z"
        status: "idle"
        version: "2.4.1"
       }
       charm: ""
       leader: true
       machine: "2"
       opened-ports: null
       public-address: "52.23.229.253"
       subordinates: null
       workload-status: {
        data: {}
        info: "ready"
        kind: ""
        life: ""
        since: "2018-08-23T14:16:31.26037557Z"
        status: "active"
        version: ""
       }
       workload-version: "16.04"
      }
     }
     workload-version: "16.04"
    }
   }
   controller-timestamp: "2018-08-23T14:46:40.334238019Z"
   machines: {
    "2": {
     agent-status: {
      data: {}
      info: "agent is not communicating with the server"
      kind: ""
      life: ""
      since: "2018-08-23T14:15:18.246038417Z"
      status: "down"
      version: "2.4.1"
     }
     constraints: "cores=1 mem=1024M"
     containers: {}
     dns-name: "52.23.229.253"
     hardware: "arch=amd64 cores=1 cpu-power=350 mem=3840M root-disk=8192M availability-zone=us-east-1b"
     has-vote: false
     id: "2"
     instance-id: "i-0f4045bcef2e06c3d"
     instance-status: {
      data: {}
      info: "running"
      kind: ""
      life: ""
      since: "2018-08-23T14:13:24.377632834Z"
      status: "running"
      version: ""
     }
     ip-addresses: [
      "52.23.229.253"
      "172.31.39.86"
      "252.39.86.1"
     ]
     jobs: [
      "JobHostUnits"
     ]
     network-interfaces: {
      eth0: {
       gateway: "172.31.32.1"
       ip-addresses: [
        "172.31.39.86"
       ]
       is-up: true
       mac-address: "12:ae:04:27:c9:3a"
      }
      fan-252: {
       ip-addresses: [
        "252.39.86.1"
       ]
       is-up: true
       mac-address: "32:86:99:8b:b8:a8"
      }
     }
     series: "xenial"
     wants-vote: false
    }
   }
   model: {
    available-version: ""
    cloud-tag: "cloud-aws"
    meter-status: {
     color: ""
     message: ""
    }
    model-status: {
     data: {}
     info: ""
     kind: ""
     life: ""
     since: "2018-08-23T14:04:02.911112932Z"
     status: "available"
     version: ""
    }
    name: "jimmmodel"
    region: "us-east-1"
    sla: "unsupported"
    type: "iaas"
    version: "2.4.1"
   }
   offers: {}
   relations: null
   remote-applications: {}
  }
 }

Note that the machine-2 status is reported as "down".

By contrast, here's the machine entry as reported by WatchAll.
The agent status is reported as "started", not "down".

 machine 2 04c4615d-fcab-4b3b-8c91-1a201cf50222 {
  model-uuid: "04c4615d-fcab-4b3b-8c91-1a201cf50222"
  id: "2"
  instance-id: "i-0f4045bcef2e06c3d"
  agent-status: {
   current: "started"
   message: ""
   since: "2018-08-23T14:15:18.246038417Z"
   version: "2.4.1"
  }
  instance-status: {
   current: "running"
   message: "running"
   since: "2018-08-23T14:13:24.377632834Z"
   version: ""
  }
  life: "alive"
  series: "xenial"
  supported-containers: [
   "lxd"
  ]
  supported-containers-known: true
  hardware-characteristics: {
   arch: "amd64"
   mem: 3840
   root-disk: 8192
   cpu-cores: 1
   cpu-power: 350
   availability-zone: "us-east-1b"
  }
  jobs: [
   "JobHostUnits"
  ]