feat: HDF5 2.0.0

[kiranshila]

Will fix #113

Changes

• Bumped the vendored version of HDF5 to 2.0.0
• Tweaked the Cmake config to follow their changes, including zlib and removing the debug prefix
• Updated the -sys crate to allow for linking against 2.0 versions
• Fixed a bad test that failed on 2.0 (the ordering of member_names is not defined)

Open issues

• Support for native-complex, I would love advise on the best way to tackle it
• All of the driver _init functions have been removed. These were part of internal APIs and apparently were not meant to be used. The Julia folks ran into this first with Initialization of file drivers , H5FD_core_init, H5FDperform_init HDFGroup/hdf5#1809. I've implemented a fix that I'm ok-ish with, but this certainly needs another set of eyes. Specifically, the MPI init is left as a TODO.

With these changes, a quick test using the -src provided crate works as well as linking against an older pre 2.0 version.

[kiranshila]

Thanks for your help on this @mulimoen

I was having the same issue of trying to figure out how to properly feature-gate the inclusion of the
H5Pset_fapl_mpio fn

[kiranshila]

And ofc there’s still the todo on how to build the driver