Changed gradient type from double[dims] to Point<dims, double> because...

Changed gradient type from double[dims] to Point<dims, double> because preferred by Paraview and because Point has already a function norm, s.t. getting the gradient magnitude doesn't require an extra function.

openfpm_numerics @ e234c028

-Subproject commit d2fe7f0b9818476e414a549a28b721b3833588d8
+Subproject commit e234c0281f2b2ff1474306c77a62462f8d0971da