Allow GPU version to compile with OpenMP

[samhatfield]

This PR permits us to build ecTrans for GPUs using OpenMP offload. It is not even remotely working yet. Here I've just done all the work necessary to run a simple test program:

PROGRAM TEST_PROGRAM

USE PARKIND1, ONLY: JPIM

IMPLICIT NONE

INTEGER(JPIM) :: TRUNC

#include "setup_trans0.h"
#include "setup_trans.h"

CALL DR_HOOK_INIT()

CALL SETUP_TRANS0(LDMPOFF=.TRUE., KPRINTLEV=2)

TRUNC = 79

CALL SETUP_TRANS(KSMAX=TRUNC, KDGL=2 * (TRUNC + 1))

END PROGRAM TEST_PROGRAM

I plan to add support for OpenMP offload gradually such that we can eventually run the full benchmark program. This PR is the first step.

I've also taken the opportunity to tidy up SETUP_TRANS.

[samhatfield]

@wdeconinck this commit looks significant, but the changes here don't affect OpenACC functionality. I just made the minimum changes to permit OpenMP-offload compilation, so I don't think this PR needs much scrutiny. Could you approve?

[PaulMullowney]

@wdeconinck @samhatfield I have 2 PRs, to this PR, that I would like to be a part of this. I have tested both on Nvidia devices using OpenACC. Results look good. I am going to check each on AMD devices with Cray.

[wdeconinck]

Thanks! I much appreciate the effort to look into OpenMP offloading again.

[PaulMullowney]

Thanks! I much appreciate the effort to look into OpenMP offloading again.

Of course! Happy to help.