Add support for NVTX to aid GPU profiling

Description

Add a wrapper around the NVTX (NVIDIA Tools Extension) profiling functions using a Fortran and C interface to the C functions as defined in the corresponding header. Also modify configure.ac and the cuda m4 macro to add the necessary header file and library.

The NVTX functions are called in profiling_in and profiling_out to create regions around all functions for which profiling is done right now. This facilitates GPU profiling because one has a timeline where it is clear at which time the code is in which function.

News snippet

Add support for NVTX to aid GPU profiling

Checklist

• I have checked that my code follows the Octopus coding standards