Add multiphase baroclinic term (multiscale rhs for baroclinic term on gpu, multiscale grad(P) on cpu, baroclinic term from a given rhs field). Add tests (test_multiphase_baroclinic for GPU seems broken)

Profile GPU-diffusion initilialisation. Fix multiresolution filter, GPU advection, redistribute op. Fix some tests.

fix small bug in poisson_fft + new version of stretching operator encompassing the resolution of the linearized stretching equation

add SFD stuff (ie: method to get the filtered solution (=the base flow) needed to perform global linear stability analysis)

Update single precision patch. Add RMI example (need further tests, fix initial condition). Improve baroclinic operator.

All tests ok, except (as usual) :
 22 - mpi_test_advec_scales (Failed)
 79 - test_advection_nullVelocity (Failed)
 81 - test_advection_randomVelocity (Failed)
 85 - test_gpu_multiresolution_filter (Failed)
 86 - mpi_test_gpu_multiresolution_filter (Failed)