Description
Currently FileDataStore.getAllIdentifiers evaluates all file eagerly and maintains the set in memory
As an optimization i.e. prevent loading of all file names in memory we can use com.google.common.io.Files#fileTreeTraverser. This is currently used in org.apache.jackrabbit.oak.spi.blob.FileBlobStore#getAllChunkIds
Same logic can be done for OakFileDataStore which extends the FileDataStore