Make hdf5 printers and build system more robust to hdf5 version differences

[anderkve]

This PR adds some modifications to Printers/ and to our CMake system, to make the GAMBIT build more robust to issues related to the API changes between different versions of the HDF5 library.

Note that we should probably wait for PR #582 to be merged first and pull that small change into this PR before merging this one.

Summary of changes:

Changes to CMake system (cmake/optional.cmake):

• Set HDF5_PREFER_PARALLEL FALSE before find_package so a parallel HDF5 install can't be silently preferred. GAMBIT's printers are serial. (MPI coordination is done on the GAMBIT side, not by the parallell HDF5 library.)
• Switched include_directories(...) to include_directories(SYSTEM ...).
• Dropped the global -DH5_USE_110_API. (When not scoped, this would affect all our build targets, not just the GAMBIT code in Printers.)
• Added a check_c_source_compiles test (GAMBIT_HDF5_USABLE) that compiles + links a small program against the discovered HDF5; if it fails, the printers get ditched with a clear
  hint to use -DHDF5_ROOT=.... This should ensure that we catch the cases where different hdf5 libraries are mixed early in the configure process, rather that giving some cryptic compile/link error deep into the GAMBIT build.
• Added a -DHDF5_ROOT=... hint to the not-found message.

Changes to Printers/ (so we no longer need -DH5_USE_110_API):

• Printers/include/gambit/Printers/printers/hdf5printer/hdf5tools.hpp: added a small set of macros that maps GAMBIT_H5L_INFO_T, GAMBIT_H5O_INFO_T, GAMBIT_H5OGET_INFO_BY_NAME(...)
  and GAMBIT_H5LITERATE to the version-specific symbols (H5Literate2/H5Oget_info_by_name3/H5O_info2_t/H5L_info2_t on 1.12+, the unversioned forms on 1.10).
• Printers/src/printers/hdf5printer/hdf5tools.cpp: group_ls no longer uses the deprecated H5Gget_objinfo/H5G_stat_t (now uses GAMBIT_H5OGET_INFO_BY_NAME + H5O_TYPE_DATASET);
  lsGroup uses GAMBIT_H5LITERATE; isDataSet uses the macros.
• Printers/src/printers/hdf5printer/hdf5_combine_tools.cpp: op_func/op_func_aux updated to use the macros

[anderkve]

This is a follow-up to what we discussed in the Core meeting today, @pstoecker, @ChrisJChang and @carlmfe. I added @carlmfe as reviewer, but if anyone else gets to it first, feel free to grab it.