Allow LIMHEL>0 in cudacpp

[valassi]

This is a followup to #950.

In 950, which is now complete (and will merged in #955 and #986), @oliviermattelaer did two things amongst others

• LIMHEL in fortran is now controlled by a runcard
• in cudacpp, the runcard sets limhel=0 by default for now (see df130b0772)

What remains to be done is to allow limhel to be different from 0 in cudacpp. This can be done with lower priority in a different PR.

[valassi]

OPS. This is a duplicate of #564. I close this and move all links to #564.