[Feature Request] Add parallel support to H5Ocopy

[jhendersonHDF]

H5Ocopy does not currently take advantage of multiple MPI processes when HDF5 is built with parallel enabled. This can cause performance problems for various reasons, such as the fact that H5Ocopy will have all processes copy the same raw data into a new dataset rather than performing a distributed collective write or simply having MPI rank 0 write the data. This discussion is for improving H5Ocopy in parallel HDF5.