When executors are lost, the status is not propagated back to zookeeper

[philwinder]

The healthcheck calls the executor lost method, but this method isn't implemented in the scheduler.

This depends on #303 and #318

Need code like the following:

        for (Protos.TaskInfo taskInfo : clusterState.getTaskList()) {
            if (taskInfo.getExecutor().getExecutorId().getValue().equals(executorId.getValue())) {
                LOGGER.debug("Found task match for " + executorId.getValue() + ". Sending status update.");
                statusUpdate(driver, Protos.TaskStatus.newBuilder().setExecutorId(executorId).setSlaveId(slaveId).setTaskId(taskInfo.getTaskId()).setState(Protos.TaskState.TASK_ERROR).build());
            }
        }

[philwinder]

Fixed in #330