Cache the repository instance that has the Hive Metastore connection

While ingesting the events to Kite dataset sink via Flume, we consistently notice the attempts to connect to the Hive metastore for every batch:

INFO hive.metastore: Trying to connect to metastore with URI thrift://<some-node>:9083

The repository instance that has the metastore connection should be cached for reuse instead of opening a new connection to the metastore for each batch.