taskdistribution works on failover mode if master is undefined, so it is not needed to have 2 methods

  Consider the last start date to not include same entry multiple times

/reviewed-on nexedi/slapos.toolbox!11

  In general, We can consider that at worst 1ms is added to
  ipv4 response, due the usage of openvpn.

  We can also consider that we accept a certain increase
  on latency for various reasons, if we are on a
  congested link. So we can consider that 10% is
  reseonable enough.

  The goal of this promise it to detect the real slowliness, rather
  them minor temporary failures, due internet oscilation.

  Add new promise to check re6st is on optimal status

  Added a promise to test if there is a packet lost when ping to an IPv4/IPv6.

  Added new promise to inspect an apache error log to detect failures if failures occurred.

@Tyagov 

Some screenshots:
![Screenshot_from_2017-03-24_11-54-11](/uploads/16f5901303b083dee2c771bcc77dd93b/Screenshot_from_2017-03-24_11-54-11.png)

/reviewed-on nexedi/slapos.toolbox!10

Until now, if json item files couldn't be read, the script was
crashing. This commit delete the corrupted files as they do not
provide any useful information.