Fix compilation errors on Intel; batched FFTs not yet enabled for Intel GPUs
Compilation on Intel GPUs needs some macro protection of some HIP constructs and routines in order to work again. Moreover, inside Intel and AMD common code blocks, the __NO_HIPKERN
macro is used together with __ONEMKL
macro to always prevent HIP kernels compilation if the executable targets Intel GPUs.
Edited by Giacomo Rossi