Fix #121

[BenWhitehead]

The executor information was not properly cleared when an executor was lost in the cluster resulting in potentially invalid executor configuration when attempting to restart an executor when the scheduler has moved.

[BenWhitehead]

Tested on dcos cluster with following steps:

1. Install cassandra dcos package install --yes cassandra
2. Wait for cassandra to come up and be healthy
3. Navigate to marathon UI and force a restart of the cassandra scheduler making sure it has moved to a new host.
4. SSH to a slave running a cassandra node
5. Terminate executor process with kill -9
6. Ensure that process is able to restart on the same node, with a new executor id.