Improve Oracle redo logs query to avoid de-duplication step

In LogMinerHelper#setRedoLogFilesForMining there is a call to get the online redo logs but the underlying query does not factor out those that have been archived, which leads to a step where we need to de-duplicate entries. This could be optimized to avoid this extra filter step. We could also possibly benefit from a bidirectional map, see https://github.com/debezium/debezium/pull/2186#discussion_r588339094