Update cassandra loading jar

We were using our own refinery-job jar, containing a dependency on the spark-cassandra-connector package. However the spark-cassandra-connector package has a bug preventing us to load non-normalized columns (columns with a dash in their name for instance). This patch updates the job to use a manually patched and released version of the spark-cassandra-connector package to unblock the situation. The long term plan is to change the cassandra schemas to only use normalized column names.