GitLab

OK, so here's a number of changes - mainly to build system - to make it
easier to use CRMC shared libraries in third party applications.

The changes are

- Use **CMAKE** `configure_file` macro instead of hand-crafted replacements
  This effects all files ending in `-HEPMCTEMPLATE` or `-HEPMC-TEMPLATE`
- In case ROOT is enable, do not set `-std=c++X` - ROOT will supply that
  option
- Sub-module `CMakeLists.txt` search for files not only in current source
  directory but also current binary directory.  This is needed for
  out-of-source builds (don't eat where you ... - you get the idea)
  so that **CMAKE** can find the generated files (from `configure_file`)
- Enforce that only _one_ **DPMJet** model version can be built
- Enforce, in case of static linking, that **PhoJet** and **DPMJet**
  cannot be built together
- In case of static linking, check if both C++ and Fortran compilers
  support the flag `-mcmodel=medium` and if so, pass that flag.  This
  is needed - otherwise we might get the error

      relocation truncated to fit: R_X86_64_32S against symbol ...

  from GCC (and I think other compilers too)
- Use

      set_property(SOURCE files APPEND PROPERTY COMPILER_DEFINITIONS ...)

  instead of setting `COMPILER_FLAGS` in a two step fashion.
- Some other minor code fixes

  - `src/models.F` Preprocessor guard around `PHO_PHIST` expanded
    to take into account if PhoJet is being built
  - `src/dpmjet/3.2019-1/DPMJET-19.1/src/phojet/PHO_SETPCOMB.f` - fix
    bug. Trying to format a `CHARACTER` array as `REAL`.

Some things to consider for the future:

- Perhaps EPOS should be treated like any other model
- Perhaps define own transform from `HEPEVT` to `HepMC3::GenEvent`
  procdure so that the size of the `HEPEVT` common block can be
  directly defined here
- ROOT output should really use the "standard" `TGenerator` output -
  that is, a `TClonesArray` of `TParticle` objects.  Additional
  information can still be stored in the output tree (separate branches)