Performance of cube file writing with MPI I/O

[Ondrej Marsalek]

Dear all,

CP2K 6.1 added MPI I/O for writing cube files, which does not seem to
be optional as far as I can tell. For our particular application on
our particular cluster, it causes a massive slowdown compared to CP2K
5.1 - over a minute vs a few seconds to write one cube file. Before I
write a detailed report, I wanted to check if some else perhaps
already ran into the problem or if it's something the developers are
aware of. Ideally, it would be possible to enable/disable this in the
input file, either globally or for cube files specifically. For now, I
am patching `pw_to_cube` in the build on our cluster to never use MPI
I/O, but that is of course far from ideal.

All the best,
Ondrej