Decouple Logging system from Datasets

Log Saver and Log Reader use Datasets and Transaction to store meta information like Kafka offsets, log file names, etc. When there is an issue starting Datasets or Transaction Service, Log Reader will not be able to startup and hence logs related to Datasets issue will not be available to users to debug.

By directly writing to HBase tables, there will be no dependency on any CDAP services, and Log Reader can startup without any other service. However, we will lose transaction rollback capability on errors.