Switch from TARGETs back to POINTERs

This reverts several changes from !5 (merged). Namely, TARGET arguments of several procedures of the Fortran interface are made POINTERs again. This probably makes the Fortran interface a bit inconsistent (i.e. some functions expect POINTERs and some other expect TARGETs) but without these changes, ICON (commit) fails (log) with the following backtrace:

[util_backtrace]: #0 0x2963ae1 - mo_exception_mp_finish_ in /fastdata/mh0156/buildbot/levante5/levante_intel_hybrid/build/externals/fortran-support/src/mo_exception.F90:243
[util_backtrace]: #1 0x628ce3 - mo_ocean_time_events_mp_init_ocean_time_events_ in /fastdata/mh0156/buildbot/levante5/levante_intel_hybrid/build/src/ocean/drivers/mo_ocean_time_events.f90:125
[util_backtrace]: #2 0x6269b4 - mo_ocean_model_mp_construct_ocean_model_ in /fastdata/mh0156/buildbot/levante5/levante_intel_hybrid/build/src/ocean/drivers/mo_ocean_model.f90:385
[util_backtrace]: #3 0x6264a6 - mo_ocean_model_mp_ocean_model_ in /fastdata/mh0156/buildbot/levante5/levante_intel_hybrid/build/src/ocean/drivers/mo_ocean_model.f90:168
[util_backtrace]: #4 0x4208ca - MAIN__ in /fastdata/mh0156/buildbot/levante5/levante_intel_hybrid/build/src/drivers/icon.f90:241
[util_backtrace]: #5 0x4202a2 - main in /fastdata/mh0156/buildbot/levante5/levante_intel_hybrid/build/experiments/test_nextGEMS/work/run_20140120T120000-20140120T235500/icon:0
[util_backtrace]: #6 0x7ffefbee7d85 - __libc_start_main in /usr/lib64/libc-2.28.so:0
[util_backtrace]: #7 0x4201ae - _start in /fastdata/mh0156/buildbot/levante5/levante_intel_hybrid/build/experiments/test_nextGEMS/work/run_20140120T120000-20140120T235500/icon:0
[util_backtrace]: #8 (nil) - ?? in ??:0

I failed to identify what exactly goes wrong there, but together with !17 (merged), this should make the master work for ICON again (see here).