We should implement optimizations for the common case where multiple VDS mappings share the same source file name and/or source dataset name. Both in memory and in the file. See this short RFC: https://github.com/HDFGroup/hdf5doc/blob/master/RFCs/HDF5_Library/VDS_Shared_Strings/VDS-Shared-Strings-Design.v2.pdf