incremental read support with clustering

As part of clustering, metadata such as hoodie_commit_time changes for records that are clustered. This is specific to SparkBulkInsertBasedRunClusteringStrategy implementation. Figure out a way to carry commit_time from original record to support incremental queries.

Also, incremental queries dont work with 'replacecommit' used by clustering HUDI-1264. Change incremental query to work for replacecommits created by Clustering.