Healthcheck improvments

[j1n6]

• Support multiple healthcheck per app
• Support TCP, HTTP, COMMAND check specified in Marathon
• Try not re-invent the wheel by getting health check information from marathon

Also related to #56 #62

[j1n6]

There are a couple of requirements where the healthcheck should/shouldn't happen:

• It shouldn't curl directly to Marathon. When upgrading marathon or restart marathon, services avalabiliity would be affected.
• It shouldn't curl directly to Bamboo. When upgrading bamboo or bamboo upgrade, they will be affected.
• It should be a daemon that is 100% up all the time, doesn't involved much change.
• When marathon is down, it should keep the last health check snapshot until Marathon is up

[nickpoorman]

When doing health checks from somewhere like Route 53 it's nice to be able to check the status of marathon. A sensible default might be to add the following to the haproxy template to get the marathon status and subsequently check HAProxy is up and routing.

frontend http-in
  ...
    acl marathon-status path_beg -i /status
    use_backend status if marathon-status

backend status
    balance leastconn
    option httpclose
    option forwardfor

    server localhost 127.0.0.1:8000