Avoid use of "magic numbers", especially when deciding behavior based on version numbers

[eflumerf]

When looking at the changes in PR #38 , I noticed a lot of if(version >= 3) constructs, and was wondering if it might be better to have a HDF5FileVersion.h or similar which:

• defines the current version as a C++ constand
• contains a set of "feature flags" which can be checked in code (e.g. FormatContainsGeoIDs, FormatContainsSourceIDs), with a constructor that takes a given version number to set them appropriately
• Uses pybind to expose these constants/flags to hdf5_dump.py to avoid having things explicitly coded there (such as detdataformats::DetID::Subdetector)